Communication method, apparatus and system

Abstract

A communication method, apparatus and system, which are applied in the field of federated learning. The method comprises: a federated learning server receiving a first request message from a first federated learning member, wherein the first request message carries a first training requirement or an intermediate training result, and the intermediate training result is used for determining the first training requirement; and then the federated learning server determining a second federated learning member from among candidate federated learning members on the basis of capability information of the candidate federated learning members and the first training requirement, wherein the second federated learning member supports the continuation of executing a training task for a first model. By means of the technical solution of the present application, during the selection of an appropriate second federated learning member to continue executing a training task, the current training situation or training requirement is taken into consideration, such that an increase in time required for the convergence of a model can be avoided, thereby ensuring the model performance.

Description

Communication methods, devices and systems

[0001] This application claims priority to Chinese Patent Application No. 202411216075.3, filed with the China National Intellectual Property Administration on August 30, 2024, entitled "Communication Method, Apparatus and System", the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of communication technology, and more specifically, to a communication method, apparatus, and system. Background Technology

[0003] Federated learning (FL), a machine learning technique in the core network, allows multiple participants (e.g., server-side network data analytics function (NWDAF) and client-side NWDAF) to train locally and interact with model parameters through secure mechanisms without exchanging training data, thus achieving collaborative training. Federated learning can effectively help multiple organizations use data and learn models while meeting user privacy, data security, and regulatory requirements.

[0004] However, if the client-side NWDAF fails to complete the model training task during federated learning, it may increase the model convergence time and lead to a decrease in model performance. Summary of the Invention

[0005] This application provides a communication method, apparatus, and system that can avoid increasing model convergence time and ensure model performance.

[0006] Firstly, a communication method is provided. This method can be executed by a federated learning server. Unless otherwise specified, the "federated learning server" in this application can refer to the communication device itself, a component in the communication device (e.g., a communication module, processor, circuit, chip, or chip system), or a logic module or software that can implement all or part of the functions of the communication device.

[0007] The method includes: receiving a first request message from a first federated learning member, the first request message including a first training requirement corresponding to a first model; determining a second federated learning member from the candidate federated learning members based on the capability information of the candidate federated learning members and the first training requirement, the capability information indicating whether the candidate federated learning member supports continuing to execute the training task of the first model, at least one of the candidate federated learning members supports continuing to execute the training task of the first model, and at least one federated learning member includes the second federated learning member.

[0008] Secondly, a communication method is provided. This method can be executed by a federated learning server. Unless otherwise specified, the "federated learning server" in this application can refer to the communication device itself, a component in the communication device (e.g., a communication module, processor, circuit, chip, or chip system), or a logic module or software that can implement all or part of the functions of the communication device.

[0009] The method includes: receiving a first request message from a first federated learning member, the first request message including intermediate training results of a first model, the intermediate training results being obtained by the first federated learning member performing a training task of the first model; determining a first training requirement corresponding to the first model based on the intermediate training results; and determining a second federated learning member from the candidate federated learning members based on the capability information of the candidate federated learning members and the first training requirement, the capability information indicating whether the candidate federated learning members support continuing to perform the training task of the first model, at least one of the candidate federated learning members supporting continuing to perform the training task of the first model, and at least one federated learning member including the second federated learning member.

[0010] Optionally, the number of first or second federated learning members may be one or more, and this application does not limit this.

[0011] Understandably, the first training requirement in this application refers to the current training requirement for the training task of the first model, and the second training requirement can be understood as the overall training requirement for the training task of the first model. The first training requirement is determined based on the second training requirement and the current training situation (e.g., the intermediate training results of the first model).

[0012] Based on the above scheme, the federated learning server, based on the first training requirement or intermediate training results of the first model carried in the received first request message, can select at least one second federated learning member from the candidate federated learning members to continue the training task of the first model, ensuring the completion of the training of the first model. That is, when selecting a suitable second federated learning member to continue the training task, the federated learning server considers the current training situation or training requirements, thus avoiding an increase in model convergence time and a decrease in model performance.

[0013] In one possible design, the first request message may also include indication information indicating that the first federated learning member is unable to continue the training task of the first model, and / or, the indication information indicating that at least one federated learning member has been found.

[0014] Based on the above scheme, by carrying indication information in the first request message, the first federated learning member can be explicitly or implicitly indicated as unable to continue the training task of the first model, and / or at least one federated learning member can be discovered, so that the federated learning server can perform the process of discovering the second federated learning member, that is, select a suitable second federated learning member according to the first training requirements carried in the first request message, so as to continue the training task of the first model.

[0015] In one possible design, the method further includes sending a first response message to the first federated learning member, the first response message including information about the second federated learning member.

[0016] For example, the information of the second federated learning member includes: the identity (ID) and / or address information of the second federated learning member.

[0017] In one possible design, the method further includes sending a second request message to a second federated learning member, the second request message requesting continued execution of the training task of the first model, the second request message including the first training requirement and the intermediate training results of the first model.

[0018] Based on the above scheme, when the federated learning server discovers a second federated learning member for continuing the training task of the first model, it can feed back the information of the second federated learning member to the first federated learning member, so that the first federated learning member can subsequently request the second federated learning member to execute the training task of the first model; or, the federated learning server may not feed back the information of the second federated learning member to the first federated learning member, and then the federated learning server can request the second federated learning member to execute the training task of the first model. This provides multiple implementation methods and is flexible.

[0019] In one possible design, the intermediate training results of the first model include at least one of the following: a temporary model obtained by performing the training task of the first model; the accuracy information of the temporary model; the accuracy information required by the training task of the first model; the training time of the temporary model; the training time required to continue performing the training task of the first model; the resource usage of the temporary model; the training resources required to continue performing the training task of the first model; or, the number of training rounds of the temporary model; the number of training rounds required to continue performing the training task of the first model.

[0020] In one possible design, the method further includes: receiving a first model trained from a second federated learning member and training information of the first model; wherein the training information of the first model includes at least one of the following: the identifier of the first model; the address information of the first model; the analysis identifier corresponding to the first model; the service identifier corresponding to the first model; the application identifier corresponding to the first model; the time of obtaining the first model; the training duration of the first model; or, the accuracy information of the first model.

[0021] Based on the above scheme, the federated learning server can obtain the trained first model and its training information. That is, the training task of the first model can be continued by the second federated learning member. This can avoid the increase in model convergence time caused by the first federated learning member being unable to continue the training task of the first model, and can ensure model performance.

[0022] In one possible design, before determining the second federated learning member from the candidate federated learning members based on the capability information of the candidate federated learning members and the first training requirements, the method further includes: obtaining the capability information of the candidate federated learning members.

[0023] For example, the capability information is also used to indicate at least one of the following: the training time supported by the candidate federated learning member; the available training time corresponding to the candidate federated learning member; the training resources supported by the candidate federated learning member; the available training resources corresponding to the candidate federated learning member; or, the number of model training epochs supported by the candidate federated learning member.

[0024] Based on the above scheme, by obtaining the capability information of the candidate federated learning members, upon receiving the first request message from the first federated learning member, a suitable second federated learning member can be selected based on the capability information of the candidate federated learning member and the current training status to continue the training task of the first model, thereby avoiding an increase in model convergence time and a decrease in model performance.

[0025] In one possible design, the first training requirement includes at least one of the following: accuracy information required to continue the training task of the first model; training time required to continue the training task of the first model; training resources required to continue the training task of the first model; or, the number of training epochs required to continue the training task of the first model.

[0026] For example, the training time required to continue performing the training task of the first model is determined based on the training time required by the training task of the first model and the training time of the temporary model obtained by performing the training task of the first model.

[0027] For example, the number of training rounds required to continue the training task of the first model is determined based on the number of training rounds required by the training task of the first model and the number of training rounds of the temporary model obtained by performing the training task of the first model.

[0028] In one possible design, before receiving the first request message from the first federated learning member, the method further includes: receiving a third request message from the model requester, the third request message being used to request the execution of a training task for the first model, the third request message including a second training requirement corresponding to the first model; determining the first federated learning member from the candidate federated learning members based on the capability information of the candidate federated learning members and the second training requirement; sending a fourth request message to the first federated learning member, the fourth request message being used to request the execution of a training task for the first model, the fourth request message including a second training requirement; wherein the first training requirement is determined based on the second training requirement and the intermediate training results of the first model.

[0029] In one possible design, the second training requirement includes at least one of the following: the analysis identifier corresponding to the first model; the service identifier corresponding to the first model; the application identifier corresponding to the first model; the time for obtaining the first model and the training duration of the first model; the accuracy information required by the training task of the first model; the training time required by the training task of the first model; the training resources required by the training task of the first model; or, the number of training rounds required by the training task of the first model.

[0030] In one possible design, the first or second federated learning member is either an enabling client or an enabling server.

[0031] Thirdly, a communication method is provided. This method can be executed by a first federated learning member. Unless otherwise specified, the "first federated learning member" in this application can refer to the communication device itself (e.g., a terminal or access network), a component in the communication device (e.g., a communication module, processor, circuit, chip, or chip system), or a logic module or software that can implement all or part of the functions of the communication device.

[0032] The method includes: determining a first training requirement corresponding to the first model based on the intermediate training result and the second training requirement when it is determined that a first federated learning member cannot continue to perform the training task of the first model, and / or when it is determined that the intermediate training result of the first model does not meet the second training requirement corresponding to the first model; sending a first request message to a federated learning server, the first request message including the first training requirement; and receiving a first response message from the federated learning server, the first response message including information of a second federated learning member, the second federated learning member supporting the continued execution of the training task of the first model.

[0033] Fourthly, a communication method is provided. This method can be executed by a first federated learning member. Unless otherwise specified, the "first federated learning member" in this application can refer to the communication device itself (e.g., a terminal or access network), a component in the communication device (e.g., a communication module, processor, circuit, chip, or chip system), or a logic module or software that can implement all or part of the functions of the communication device.

[0034] The method includes: when it is determined that a first federated learning member cannot continue to execute the training task of a first model, and / or when it is determined that the intermediate training result of the first model does not meet the second training requirement corresponding to the first model, sending a first request message to a federated learning server, wherein the first request message includes the intermediate training result of the first model, the intermediate training result is used to determine the first training requirement corresponding to the first model, and the intermediate training result is obtained by the first federated learning member executing the training task of the first model; receiving a first response message from the federated learning server, the first response message including information of a second federated learning member, the second federated learning member supporting the continued execution of the training task of the first model.

[0035] Based on the above scheme, if a first federated learning member determines that it cannot continue training the first model, and / or the intermediate training results of the first model do not meet the second training requirements corresponding to the first model, it can request the federated learning server to discover at least one federated learning member, such as a second federated learning member, to continue training the first model. Based on the first training requirements carried in the received first request message or the intermediate training results of the first model, the federated learning server can determine a second federated learning member from the candidate federated learning members to continue training the first model, ensuring the completion of the first model's training. That is, when selecting a suitable second federated learning member to continue training, the federated learning server considers the current training situation or training requirements, avoiding an increase in model convergence time and a decrease in model performance.

[0036] In one possible design, the method further includes sending a second request message to a second federated learning member, the second request message requesting continued execution of the training task of the first model, the second request message including the first training requirement and intermediate training results.

[0037] In one possible design, the first request message may also include indication information indicating that the first federated learning member is unable to continue the training task of the first model, and / or, the indication information indicating that at least one federated learning member has been found.

[0038] In one possible design, before sending the first request message to the federated learning server, the method further includes sending capability information of the first federated learning member to the federated learning server, the capability information indicating that the first federated learning member supports the training task of the first model.

[0039] In one possible design, the capability information is also used to indicate at least one of the following: the training time supported by the first federated learning member; the available training time corresponding to the first federated learning member; the training resources supported by the first federated learning member; the available training resources corresponding to the first federated learning member; or, the number of model training epochs supported by the first federated learning member.

[0040] In one possible design, before determining the first training requirement corresponding to the first model based on the intermediate training results and the second training requirement, or before sending the first request message to the federated learning server, the method further includes: receiving a fourth request message from the federated learning server, the fourth request message being used to request the execution of the training task of the first model, the fourth request message including the second training requirement; executing the training task of the first model according to the second training requirement to obtain intermediate training results.

[0041] In one possible design, the first training requirement includes at least one of the following: the accuracy information required to continue the training task of the first model; the training time required to continue the training task of the first model; the training resources required to continue the training task of the first model; or, the number of training epochs required to continue the training task of the first model.

[0042] In one possible design, the second training requirement includes at least one of the following: the analysis identifier corresponding to the first model; the service identifier corresponding to the first model; the application identifier corresponding to the first model; the time for acquiring the first model; the training duration of the first model; or, the accuracy information required by the training task of the first model; the training time required by the training task of the first model; the training resources required by the training task of the first model; or, the number of training rounds required by the training task of the first model.

[0043] In one possible design, the intermediate training results include at least one of the following: a temporary model obtained by performing the training task of the first model; the accuracy information of the temporary model; the accuracy information required by the training task of the first model; the training time of the temporary model; the training time required to continue performing the training task of the first model; the resource usage of the temporary model; the training resources required to continue performing the training task of the first model; the number of training rounds of the temporary model; or, the number of training rounds required to continue performing the training task of the first model.

[0044] The beneficial effects corresponding to the third or fourth aspect mentioned above can be referred to the relevant descriptions of the beneficial effects corresponding to the first or second aspect, and will not be repeated here.

[0045] Fifthly, a communication method is provided. This method can be executed by a second federated learning member. Unless otherwise specified, the "second federated learning member" in this application can refer to the communication device itself (e.g., a terminal or access network), a component in the communication device (e.g., a communication module, processor, circuit, chip, or chip system), or a logic module or software that can implement all or part of the functions of the communication device.

[0046] The method includes: receiving a second request message from a first federated learning member or a federated learning server, the second request message being used to request continued execution of the training task of the first model, the second request message including intermediate training results of the first model and a first training requirement corresponding to the first model; continuing to execute the training task according to the intermediate training results and the first training requirement to obtain the first model and the training information of the first model.

[0047] Furthermore, the first model and its training information are sent to the federated learning server.

[0048] Based on the above scheme, the second federated learning member can continue to perform training tasks in a targeted manner based on the intermediate training results and the first training requirements carried in the received second request message, obtain the first model and the training information of the first model, ensure the completion of the training of the first model, avoid the increase of model convergence time, and avoid the degradation of model performance.

[0049] In one possible design, before continuing the training task based on intermediate training results and the first training requirements, the method further includes: determining whether to continue the training task based on the capability information of the second federated learning member and the first training requirements.

[0050] In one possible design, before receiving a second request message from the first federated learning member or the federated learning server, the method further includes: sending capability information of the second federated learning member to the federated learning server, the capability information indicating that the second federated learning member supports continuing the training task of the first model.

[0051] In one possible design, the capability information is also used to indicate at least one of the following: the training time supported by the second federated learning member; the available training time corresponding to the second federated learning member; the training resources supported by the second federated learning member; the available training resources corresponding to the second federated learning member; or, the number of model training epochs supported by the second federated learning member.

[0052] In one possible design, the intermediate training results include at least one of the following: a temporary model obtained by performing the training task of the first model; the accuracy information of the temporary model; the accuracy information required by the training task of the first model; the training time of the temporary model; the training time required to continue performing the training task of the first model; the resource usage of the temporary model; the training resources required to continue performing the training task of the first model; the number of training rounds of the temporary model; or, the number of training rounds required to continue performing the training task of the first model.

[0053] In one possible design, the training information of the first model includes at least one of the following: the identifier of the first model; the address information of the first model; the analysis identifier corresponding to the first model; the service identifier corresponding to the first model; the application identifier corresponding to the first model; the time of acquiring the first model; the training duration of the first model; or the accuracy information of the first model.

[0054] In one possible design, the first training requirement includes at least one of the following: the accuracy information required to continue the training task of the first model; the training time required to continue the training task of the first model; the training resources required to continue the training task of the first model; or, the number of training epochs required to continue the training task of the first model.

[0055] In one possible design, the first or second federated learning member is either an enabling client or an enabling server.

[0056] The beneficial effects corresponding to the fifth aspect mentioned above can be referred to the relevant descriptions of the first or second aspects, and will not be repeated here.

[0057] Sixthly, a communication method is provided. This method can be executed by a federated learning server. Unless otherwise specified, the "federated learning server" in this application can refer to the communication device itself, a component in the communication device (e.g., a communication module, processor, circuit, chip, or chip system), or a logic module or software that can implement all or part of the functions of the communication device.

[0058] The method includes: receiving a third request message from a model requester, the third request message being used to request the execution of a training task of a first model, the third request message including a second training requirement corresponding to the first model; determining a first federated learning member and a second federated learning member from the candidate federated learning members based on the capability information of the candidate federated learning members and the second training requirement, the capability information indicating whether the candidate federated learning members support continuing to execute the training task of the first model, at least one of the candidate federated learning members supporting continuing to execute the training task of the first model, the at least one federated learning member including a second federated learning member, the second federated learning member being a candidate federated learning member for executing the training task of the first model; and sending a fourth request message to the first federated learning member, the fourth request message being used to request the execution of the training task of the first model, the fourth request message including the second training requirement corresponding to the first model and information of the second federated learning member.

[0059] Understandably, the second federated learning member is an alternative (or standby) federated learning member used to perform the training task of the first model. Specifically, the second federated learning member can be seen as an alternative federated learning member to the first federated learning member. The alternative (or standby) federated learning member can be understood as follows: the second federated learning member can continue to perform the training task of the first model when the first federated learning member is unable to do so, ensuring that the first model can be trained successfully. In other words, the second federated learning member supports the continued execution of the training task of the first model, or in other words, the second federated learning member has the ability to continue the training task of the first model.

[0060] Based on the above scheme, the federated learning server can select a first federated learning member and a second federated learning member from the candidate federated learning members according to their capability information and the second training requirements. The second federated learning member supports continuing the training task of the first model, especially when the first federated learning member cannot continue the training task, and / or the intermediate training results of the first model do not meet the second training requirements, ensuring the completion of the first model's training. That is, when selecting a federated learning member to execute the training task, the federated learning server can simultaneously determine the first and second federated learning members. This allows the server to request the second federated learning member to continue the training task if the first federated learning member subsequently becomes unable to do so. In other words, the first federated learning member and the federated learning server do not need subsequent signaling interactions (e.g., the first federated learning member requests the federated learning server to discover at least one federated learning member), which avoids increased model convergence time, prevents model performance degradation, and improves model training efficiency.

[0061] In one possible design, before determining the first and second federated learning members from the candidate federated learning members based on their capability information and the second training requirements, the method further includes: obtaining the capability information of the candidate federated learning members.

[0062] In one possible design, the capability information is also used to indicate at least one of the following: the training time supported by the candidate federated learning member; the available training time corresponding to the candidate federated learning member; the training resources supported by the candidate federated learning member; the available training resources corresponding to the candidate federated learning member; or, the number of model training epochs supported by the candidate federated learning member.

[0063] In one possible design, the second training requirement includes at least one of the following: the analysis identifier corresponding to the first model; the service identifier corresponding to the first model; the application identifier corresponding to the first model; the time for acquiring the first model; the training duration of the first model; the accuracy information required by the training task of the first model; the training time required by the training task of the first model; the training resources required by the training task of the first model; or the number of training rounds required by the training task of the first model.

[0064] In one possible design, the first or second federated learning member is either an enabling client or an enabling server.

[0065] The beneficial effects corresponding to the sixth aspect mentioned above can be referred to the relevant descriptions in the first or second aspects, and will not be repeated here.

[0066] Seventhly, a communication method is provided. This method can be executed by a first federated learning member. Unless otherwise specified, the "first federated learning member" in this application can refer to the communication device itself (e.g., a terminal or access network), a component in the communication device (e.g., a communication module, processor, circuit, chip, or chip system), or a logic module or software that can implement all or part of the functions of the communication device.

[0067] The method includes: obtaining information about a second federated learning member, wherein the second federated learning member supports continuing the training task of the first model; determining a first training requirement corresponding to the first model based on the intermediate training result and the second training requirement when it is determined that the first federated learning member cannot continue the training task of the first model, and / or when it is determined that the intermediate training result of the first model does not meet the second training requirement corresponding to the first model; and sending a second request message to the second federated learning member, wherein the second request message is used to request to continue the training task, and the second request message includes the first training requirement and the intermediate training result.

[0068] Based on the above scheme, if the first federated learning member is unable to continue training the first model, and / or the intermediate training results of the first model do not meet the second training requirements, the first federated learning member can request the second federated learning member to continue training the first model based on the information obtained from the second federated learning member, ensuring the completion of the first model's training. In other words, at this point, the first federated learning member and the federated learning server do not need to engage in subsequent signaling interactions (e.g., the first federated learning member requests the federated learning server to discover at least one federated learning member), which can improve the efficiency of model training and ensure model performance.

[0069] In one possible design, obtaining information about the second federated learning member includes: receiving a fourth request message from the federated learning server, the fourth request message being used to request the execution of a training task for the first model, the fourth request message including a second training requirement and information about the second federated learning member; the method further includes: executing a training task according to the second training requirement to obtain intermediate training results.

[0070] In one possible design, before obtaining information about the second federated learning member, the method further includes sending capability information of the first federated learning member to the federated learning server, the capability information indicating that the first federated learning member supports the training task of the first model.

[0071] In one possible design, the capability information is also used to indicate at least one of the following: the training time supported by the first federated learning member; the available training time corresponding to the first federated learning member; the training resources supported by the first federated learning member; the available training resources corresponding to the first federated learning member; or, the number of model training epochs supported by the first federated learning member.

[0072] In one possible design, the first training requirement includes at least one of the following: the accuracy information required to continue the training task of the first model; the training time required to continue the training task of the first model; the training resources required to continue the training task of the first model; or, the number of training epochs required to continue the training task of the first model.

[0073] In one possible design, the second training requirement includes at least one of the following: the analysis identifier corresponding to the first model; the service identifier corresponding to the first model; the application identifier corresponding to the first model; the time for acquiring the first model; the training duration of the first model; the accuracy information required by the training task of the first model; the training time required by the training task of the first model; the training resources required by the training task of the first model; or, the number of training rounds required by the training task of the first model.

[0074] In one possible design, the intermediate training results include at least one of the following: a temporary model obtained by performing the training task of the first model; the accuracy information of the temporary model; the accuracy information required by the training task of the first model; the training time of the temporary model; the training time required to continue performing the training task of the first model; the resource usage of the temporary model; the training resources required to continue performing the training task of the first model; the number of training rounds of the temporary model; or, the number of training rounds required to continue performing the training task of the first model.

[0075] In one possible design, determining the first training requirement corresponding to the first model based on intermediate training results and second training requirements includes: determining the training time required to continue executing the training task of the first model based on the training time required by the training task of the first model and the training time of the temporary model obtained by executing the training task of the first model; and / or determining the number of training rounds required to continue executing the training task of the first model based on the number of training rounds required by the training task of the first model and the number of training rounds of the temporary model obtained by executing the training task of the first model.

[0076] In one possible design, the first or second federated learning member is either an enabling client or an enabling server.

[0077] The beneficial effects corresponding to the seventh aspect mentioned above can be referred to the relevant description in the sixth aspect, and will not be repeated here.

[0078] Eighthly, a communication device is provided, which has the functions of the first aspect described above. For example, the communication device includes modules, units, or means corresponding to the operations involved in the first aspect. The modules, units, or means can be implemented by software, hardware, or a combination of software and hardware.

[0079] For example, the communication device may be the aforementioned federated learning server, or a module or unit (e.g., a chip, a chip system, or a circuit) corresponding to the methods, operations, steps, or actions described in the first aspect above.

[0080] In one possible implementation, the communication device includes a transceiver unit (or communication module) and a processing unit (or processing module) connected to the transceiver unit.

[0081] For example, the transceiver unit is configured to receive a first request message from a first federated learning member, the first request message including a first training requirement corresponding to the first model; the processing unit is configured to determine a second federated learning member from the candidate federated learning members based on the capability information of the candidate federated learning members and the first training requirement, the capability information indicating whether the candidate federated learning member supports continuing to execute the training task of the first model, at least one of the candidate federated learning members supports continuing to execute the training task of the first model, and at least one federated learning member includes the second federated learning member.

[0082] Ninthly, a communication device is provided, which has the functions of the second aspect above. For example, the communication device includes modules, units or means corresponding to the operations involved in the second aspect above. The modules, units or means can be implemented by software, or by hardware, or by a combination of software and hardware.

[0083] For example, the communication device may be the aforementioned federated learning server, or a module or unit (e.g., a chip, a chip system, or a circuit) corresponding to the methods, operations, steps, or actions described in the second aspect above.

[0084] In one possible implementation, the communication device includes a transceiver unit (or communication module) and a processing unit (or processing module) connected to the transceiver unit.

[0085] For example, the transceiver unit is configured to receive a first request message from a first federated learning member, the first request message including intermediate training results of the first model, the intermediate training results being obtained by the first federated learning member performing the training task of the first model; the processing unit is configured to determine a first training requirement corresponding to the first model based on the intermediate training results; the processing unit is further configured to determine a second federated learning member from the candidate federated learning members based on the capability information of the candidate federated learning members and the first training requirement, the capability information indicating whether the candidate federated learning members support continuing to perform the training task of the first model, at least one of the candidate federated learning members supports continuing to perform the training task of the first model, and at least one federated learning member includes the second federated learning member.

[0086] In a tenth aspect, a communication device is provided, which has the functions of the third aspect described above. For example, the communication device includes modules, units, or means corresponding to the operations involved in the third aspect described above. The modules, units, or means can be implemented by software, by hardware, or by a combination of software and hardware.

[0087] For example, the communication device may be the first federated learning member mentioned above, such as a module or unit (e.g., a chip, a chip system, or a circuit) that corresponds one-to-one with the method, operation, step, or action described in the third aspect above.

[0088] In one possible implementation, the communication device includes a transceiver unit (or communication module) and a processing unit (or processing module) connected to the transceiver unit.

[0089] For example, when it is determined that the first federated learning member cannot continue to perform the training task of the first model, and / or when it is determined that the intermediate training result of the first model does not meet the second training requirement corresponding to the first model, the processing unit is used to determine the first training requirement corresponding to the first model based on the intermediate training result and the second training requirement, wherein the intermediate training result is obtained by the first federated learning member performing the training task of the first model; the transceiver unit is used to send a first request message to the federated learning server, the first request message including the first training requirement; the transceiver unit is also used to receive a first response message from the federated learning server, the first response message including information of the second federated learning member, the second federated learning member supporting the continued execution of the training task of the first model.

[0090] In the eleventh aspect, a communication device is provided, which has the functions of the fourth aspect above. For example, the communication device includes modules, units or means corresponding to the operations involved in the fourth aspect above. The modules, units or means can be implemented by software, or by hardware, or by a combination of software and hardware.

[0091] For example, the communication device may be the first federated learning member mentioned above, such as a module or unit (e.g., a chip, a chip system, or a circuit) that corresponds one-to-one with the method, operation, step, or action described in the fourth aspect above.

[0092] In one possible implementation, the communication device includes a transceiver unit (or communication module) and a processing unit (or processing module) connected to the transceiver unit.

[0093] For example, if it is determined that the first federated learning member cannot continue to perform the training task of the first model, and / or it is determined that the intermediate training result of the first model does not meet the second training requirement corresponding to the first model, the transceiver unit is configured to send a first request message to the federated learning server, wherein the first request message includes the intermediate training result of the first model, the intermediate training result is used to determine the first training requirement corresponding to the first model, and the intermediate training result is obtained by the first federated learning member performing the training task of the first model; the transceiver unit is further configured to receive a first response message from the federated learning server, the first response message includes information about the second federated learning member, and the second federated learning member supports continuing to perform the training task of the first model.

[0094] In the twelfth aspect, a communication device is provided, which has the functions of the fifth aspect above. For example, the communication device includes modules, units or means corresponding to the operations involved in the fifth aspect above. The modules, units or means can be implemented by software, or by hardware, or by a combination of software and hardware.

[0095] For example, the communication device may be the second federated learning member mentioned above, such as a module or unit (e.g., a chip, a chip system, or a circuit) that corresponds one-to-one with the method, operation, step, or action described in the fifth aspect above.

[0096] In one possible implementation, the communication device includes a transceiver unit (or communication module) and a processing unit (or processing module) connected to the transceiver unit.

[0097] For example, the transceiver unit is configured to receive a second request message from a first federated learning member or a federated learning server, the second request message being used to request continued execution of the training task of the first model, the second request message including intermediate training results of the first model and a first training requirement corresponding to the first model; the processing unit is configured to continue execution of the training task according to the intermediate training results and the first training requirement, to obtain the first model and the training information of the first model.

[0098] The transceiver unit is also used to send the first model and its training information to the federated learning server.

[0099] In the thirteenth aspect, a communication device is provided, which has the functions of the sixth aspect above. For example, the communication device includes modules, units or means corresponding to the operations involved in the sixth aspect above. The modules, units or means can be implemented by software, or by hardware, or by a combination of software and hardware.

[0100] For example, the communication device may be the aforementioned federated learning server, or a module or unit (e.g., a chip, a chip system, or a circuit) corresponding to the methods, operations, steps, or actions described in the sixth aspect above.

[0101] In one possible implementation, the communication device includes a transceiver unit (or communication module) and a processing unit (or processing module) connected to the transceiver unit.

[0102] For example, the transceiver unit is configured to receive a third request message from a model requester, the third request message being used to request the execution of a training task for the first model, the third request message including a second training requirement corresponding to the first model; the processing unit is configured to determine a first federated learning member and a second federated learning member from the candidate federated learning members based on the capability information of the candidate federated learning members and the second training requirement, the capability information indicating whether the candidate federated learning member supports continuing to execute the training task for the first model, at least one of the candidate federated learning members supports continuing to execute the training task for the first model, the at least one federated learning member including a second federated learning member, the second federated learning member being a candidate federated learning member for executing the training task for the first model; the transceiver unit is further configured to send a fourth request message to the first federated learning member, the fourth request message being used to request the execution of the training task for the first model, the fourth request message including the second training requirement corresponding to the first model and information of the second federated learning member.

[0103] In the fourteenth aspect, a communication device is provided, which has the functions of the seventh aspect above. For example, the communication device includes modules, units or means corresponding to the operations involved in the seventh aspect above. The modules, units or means can be implemented by software, or by hardware, or by a combination of software and hardware.

[0104] For example, the communication device may be the first federated learning member mentioned above, such as a module or unit (e.g., a chip, a chip system, or a circuit) that corresponds one-to-one with the method, operation, step, or action described in the seventh aspect above.

[0105] In one possible implementation, the communication device includes a transceiver unit (or communication module) and a processing unit (or processing module) connected to the transceiver unit.

[0106] For example, the processing unit is configured to obtain information about the second federated learning member, which supports continuing the training task of the first model; if it is determined that the first federated learning member cannot continue the training task of the first model, and / or if it is determined that the intermediate training results of the first model do not meet the second training requirements corresponding to the first model, the processing unit is configured to determine the first training requirements corresponding to the first model based on the intermediate training results and the second training requirements; the sending and receiving unit is configured to send a second request message to the second federated learning member, the second request message being used to request continued execution of the training task, the second request message including the first training requirements and the intermediate training results.

[0107] In a fifteenth aspect, a communication device is provided. The communication device may be the aforementioned federated learning server, a first federated learning member, or a second federated learning member. The communication device includes a transceiver, a processor, and a memory. The processor controls the transceiver to transmit and receive signals, the memory stores a computer program, and the processor retrieves and runs the computer program from the memory, causing the communication device to perform the methods in any of the possible implementations of the first to seventh aspects.

[0108] Optionally, there may be one or more processors and one or more memories.

[0109] Alternatively, the memory can be integrated with the processor, or the memory can be set up separately from the processor.

[0110] Optionally, the transceiver includes a transmitter and a receiver.

[0111] In a sixteenth aspect, a communication device is provided, the communication device including one or more processors, the one or more processors being configured to execute a computer program or instructions, which, when executed, cause the communication device to implement the methods in any possible design or implementation of the first to seventh aspects described above.

[0112] Optionally, the communication device further includes a memory for storing part or all of the computer program or instructions that implement the functions involved in the first to third aspects described above.

[0113] Optionally, the communication device further includes an interface circuit, through which the processor communicates with other devices or components.

[0114] In a seventeenth aspect, a communication system is provided. The communication system includes a federated learning server and / or a first federated learning member, wherein the federated learning server is used to execute the methods in any of the possible implementations of the first, second, or sixth aspects described above, and the first federated learning member is used to execute the methods in any of the possible implementations of the third, fourth, or seventh aspects described above.

[0115] Optionally, the communication system further includes a second federated learning member, wherein the second federated learning member is used to perform the method in any of the possible implementations of the fifth aspect described above.

[0116] Eighteenthly, a computer-readable storage medium is provided. This computer-readable storage medium stores computer program code or instructions to cause the methods in any of the possible implementations of the first to seventh aspects to be executed, for example, when a computer reads and executes the computer program code or instructions, causing the methods in any of the possible implementations of the first to seventh aspects to be implemented.

[0117] In a nineteenth aspect, a computer program product is provided. The computer program product includes computer program code or instructions that cause the methods in any of the possible implementations of the first to seventh aspects to be implemented. For example, when a computer reads and executes the computer program product, the methods in any of the possible implementations of the first to third aspects are implemented.

[0118] In a twentieth aspect, a computer program is provided. When the computer program is run, it causes the methods in any of the possible implementations of the first to seventh aspects to be implemented.

[0119] It should be understood that the beneficial effects of aspects eight through twentieth above can be referenced from aspects one through seven above and any possible implementation thereof, and will not be elaborated here. Attached Figure Description

[0120] Figures 1 and 2 are schematic diagrams of a communication system applicable to embodiments of this application;

[0121] Figure 3 is a flowchart illustrating a federated learning execution method;

[0122] Figure 4 is a flowchart illustrating a method for transmitting intermediate models;

[0123] Figure 5 is a flowchart illustrating a communication method provided in an embodiment of this application;

[0124] Figure 6 is a flowchart illustrating another communication method provided in an embodiment of this application;

[0125] Figure 7 is a flowchart illustrating another communication method provided in an embodiment of this application;

[0126] Figure 8 is a flowchart illustrating another communication method provided in an embodiment of this application;

[0127] Figure 9 is a flowchart illustrating another communication method provided in an embodiment of this application;

[0128] Figure 10 is a schematic block diagram of a communication device provided in an embodiment of this application;

[0129] Figure 11 is a schematic block diagram of another communication device provided in an embodiment of this application. Detailed Implementation

[0130] The technical solutions in this application will now be described with reference to the accompanying drawings.

[0131] To facilitate understanding of the embodiments of this application, the following points are explained.

[0132] (1) In this application, unless otherwise specified or logically conflicting, the terms and / or descriptions of different embodiments are consistent and can be referenced by each other. The technical features of different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0133] (2) In this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, or B exists alone, where A and B can be singular or plural. In the textual description of this application, the character " / " generally indicates that the related objects before and after are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, and c can mean: a, or, b, or, c, or, a and b, or, a and c, or, b and c, or, a, b, and c. Where a, b, and c can be single or multiple.

[0134] (3) In this application, the terms "first," "second," and various numerical designations are used for convenience of description and are not intended to limit the scope of the embodiments of this application. For example, they are used to distinguish different messages, rather than to describe a specific order or sequence. It should be understood that such descriptions can be interchanged where appropriate to describe solutions other than those in the embodiments of this application.

[0135] (4) In this application, “instruction” or “for instruction” can include both direct instruction and indirect instruction. When describing an instruction as being used to instruct A, it can include whether the instruction directly instructs A or indirectly instructs A, but does not necessarily mean that the instruction carries A.

[0136] The indication methods involved in the embodiments of this application should be understood to cover various methods that enable the party to be indicated to obtain the information to be indicated. The information to be indicated can be sent as a whole or divided into multiple sub-information and sent separately. Moreover, the sending period and / or sending time of these sub-information can be the same or different. This application does not limit the sending method, for example.

[0137] The "instruction information" in the embodiments of this application can be an explicit instruction, that is, a direct instruction through signaling, or an instruction obtained by combining other rules or parameters with the parameters indicated by the signaling, or by deduction. It can also be an implicit instruction, that is, an instruction obtained based on rules or relationships, or based on other parameters, or by deduction. This application does not specifically limit it in this regard.

[0138] (5) In this application, "protocol" may refer to a standard protocol in the field of communications, for example, it may include (5) thThis application does not limit the scope of protocols such as generation (5G), new radio (NR), and related protocols applied in future communication systems. "Predefined" may include predefined terms, such as protocol definitions. "Preconfiguration" can be achieved by pre-storing corresponding codes, tables, or other means that can be used to indicate relevant information in the device; this application does not limit the implementation method.

[0139] (6) In this application, “message”, “information”, “signal” or “information element (IE)” can be used interchangeably. There are no restrictions on the name of the message or information, as long as it can achieve the corresponding function.

[0140] "Sending information to XX (device)" can be understood as the destination of the information being that device. This can include sending information directly or indirectly to that device. "Receiving information from XX (device), or receiving information from XX (device)" can be understood as the source of the information being that device. This can include receiving information directly or indirectly from that device. Information may undergo necessary processing between the source and destination, such as format changes, but the destination can understand the valid information from the source. Similar expressions in this application can be interpreted similarly and will not be elaborated upon here.

[0141] "Communication" can also be described as "communication," "information transmission," "data processing," etc. "Transmission" includes "sending" and "receiving." "Transmission" can be described as "output." "Sending" can also be understood as the "output" of a chip interface, and "receiving" can be understood as the "input" of a chip interface. In other words, "sending" or "receiving" can occur between devices, for example, between network devices and terminal devices via an air interface. "Sending" or "receiving" can also occur within a device, for example, between components, modules, chips, software modules, or hardware modules within a device via a bus, wiring, or interface.

[0142] (7) In this application, the words “exemplarily”, “for example”, etc., are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as an “example” in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word “example” is intended to present the concept in a specific manner. In the embodiments of this application, “of,” “corresponding, relevant,” “correspondingly,” and “associate” may sometimes be used interchangeably, and it should be noted that their intended meanings are consistent when their distinctions are not emphasized.

[0143] (8) In this application, the configuration can be signaling configuration or can be described as configuration signaling. For example, signaling configuration includes configuration using signaling sent by the base station, which can be radio resource control (RRC) messages, downlink control information (DCI) messages, or system information blocks (SIBs). Optionally, the signaling configuration can also be configured to the terminal device by pre-configured signaling, or configured to the terminal device through pre-configuration. Here, pre-configuration means defining or configuring the values ​​of corresponding parameters in advance in the form of a protocol, and storing them in the terminal device when communicating with the terminal device. The pre-configured messages can be modified or updated when the terminal device is connected to the network.

[0144] The technical solutions provided in this application can be applied to various communication systems, such as: new radio (NR) systems, long term evolution (LTE) systems, LTE frequency division duplex (FDD) systems, LTE time division duplex (TDD) systems, etc. The technical solutions provided in this application can also be applied to device-to-device (D2D) communication, vehicle-to-everything (V2X) communication, machine-to-machine (M2M) communication, machine-type communication (MTC), and Internet of Things (IoT) communication systems or other communication systems.

[0145] In communication systems, the portion operated by the operator can be called a public land mobile network (PLMN), or operator network, etc. A PLMN is a network established and operated to provide terrestrial mobile communication services to the public; it is primarily a public network where mobile network operators (MNOs) provide mobile broadband access services to users. The PLMN described in this application embodiment can specifically be a network conforming to the 3rd Generation Partnership Project (3GPP) standard, or simply a 3GPP network. 3GPP networks typically include, but are not limited to, 5th-generation (5G) networks, 4th-generation (4G) networks, and other future communication systems, such as 6th-generation (6G) networks.

[0146] For ease of description, this application will use PLMN or 5G networks as examples in its embodiments.

[0147] Figure 1 is a schematic diagram of a network architecture 100, taking the 5G network architecture based on service-based architecture (SBA) in a non-roaming scenario as defined in the 3GPP standardization process as an example. As shown in Figure 1, this network architecture can include a terminal equipment part, a data network (DN) part, and a carrier network PLMN part. Among them, the carrier network PLMN part can include, but is not limited to, the radio access network (RAN)120 and the core network (CN) part.

[0148] The functions of each network element are briefly explained below.

[0149] The terminal equipment portion may include terminal equipment 110, which is a device that provides voice and / or data connectivity to a user. Terminal equipment 110 may also be referred to as user equipment (UE). In this application, terminal equipment 110 is a device with wireless transceiver capabilities, capable of communicating with one or more core network (CN) devices via access network equipment (or access devices) in the (wireless) access network (RAN) 120. Terminal equipment 110 may also be referred to as an access terminal, terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, user agent, or user device, etc. Terminal equipment 110 may be a cellular phone, cordless phone, session initiation protocol (SIP) phone, smartphone, mobile phone, wireless local loop (WLL) station, personal digital assistant (PDA), etc. Alternatively, terminal device 110 can also be a handheld device with wireless communication capabilities, a computing device or other device connected to a wireless modem, an in-vehicle device, a wearable device, a drone device, or a terminal in the Internet of Things (IoT), the Internet of Vehicles (IoV), a terminal in any form in a 5G network or future network, a relay user equipment, or a terminal in a future evolved 6G network, etc. Among these, relay user equipment can be, for example, a 5G residential gateway (RG). For example, terminal device 110 can be a virtual reality (VR) terminal, an augmented reality (AR) terminal, a wireless terminal in industrial control, a wireless terminal in autonomous driving, a wireless terminal in telemedicine, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home, etc. Here, "terminal device" refers to a 3GPP terminal. This application does not limit the type or category of terminal device. For ease of explanation, this application will subsequently use "UE" to refer to the terminal device as an example.

[0150] (R)AN 120 may include one or more access network elements or access network devices. The interface between the access network device and the terminal device may be a Uu interface (or air interface, i.e., the messages exchanged between the access network device and the terminal device may be called air interface messages). In future communication, the interface name may remain unchanged or may be replaced by other names, and this application does not limit this. (R)AN 120 is a device that provides wireless communication functions for terminal device 110, and can connect the terminal device to a node or device of a wireless network, and may also be called a network device. The aforementioned RAN may be a 3GPP-related cellular system, such as a 5G mobile communication system, or a future-oriented evolution system (such as a 6G mobile communication system). RAN may also be an open radio access network (open RAN, O-RAN or ORAN), a cloud radio access network (CRAN), or a wireless fidelity (WiFi) system. (R)AN 120 can be regarded as a sub-network of the operator network, and is the implementation system between the service node and the terminal device 110 in the operator network. For example, terminal device 110 can connect to a service node in the operator's network via (R)AN 120 to obtain the services provided by that service node. (R)AN 120 includes, but is not limited to: next-generation node base station (gNB) in 5G systems, evolved node B (eNB) in LTE, radio network controller (RNC), node B (NB), base station controller (BSC), base transceiver station (BTS), home base station (e.g., home evolved node B, or home node B (HNB), base band unit (BBU), transmitting and receiving point (TRP), transmitting point (TP), small cell equipment, mobile switching center, or network equipment in future networks. Access network equipment can also be modules or units that perform base station functions, such as centralized units (CU) and distributed units (DU). This application does not limit the specific technologies or equipment forms used in the access network equipment. In systems employing different wireless access technologies, the names of devices with access network functionality may differ.For ease of description, in all embodiments of this application, the apparatus that provides wireless communication functions for terminal device 110 is collectively referred to as access network equipment or simply RAN. It should be understood that this document does not limit the specific type of access network equipment.

[0151] In different systems, CU (including open CU-CP (O-CU-CP) and open CU-UP (O-CU-UP), DU, or RU may have different names, but those skilled in the art will understand their meaning. For example, in an ORAN system, CU can also be called an open central unit (O-CU), DU can also be called an open distributed unit (O-DU), CU-CP can also be called O-CU-CP, CU-UP can also be called O-CU-UP, and RU can also be called an open radio unit (O-RU). For ease of description, this application uses CU, CU-CP, CU-UP, DU, and RU as examples. Any of the CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software and hardware modules.

[0152] Terminal device 110 or (R)AN 120 can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; it can also be deployed on water (e.g., on ships); and it can also be deployed in the air (e.g., on airplanes, balloons, and satellites).

[0153] The CN part may include, but is not limited to, the following network functions (NFs): User plane function (UPF)130, Network exposure function (NEF)131, Network function repository function (NRF)132, Policy control function (PCF)133, Unified data management (UDM)134, Unified data repository (UDR)135, Network data analytics function (NWDAF)136, Application function (AF)141, Authentication server function (AUSF)137, Access and mobility management function (AMF)138, Session management function (SMF)139.

[0154] Data network DN 140, also known as packet data network (PDN), is typically a network located outside the carrier's network, such as a third-party network.

[0155] The following is a brief explanation of the NF functions included in CN.

[0156] 1. The UPF 130 is a gateway provided by the operator, serving as the gateway for communication between the operator's network and the DN 140. UPF130 network functions include user plane-related functions such as packet routing and transmission, packet inspection, service usage reporting, Quality of Service (QoS) processing, lawful operation monitoring, uplink packet inspection, and downlink packet storage.

[0157] 2. NEF 131 is a control plane function provided by the operator, which mainly enables third parties to use the services provided by the network, supports the network to open its capabilities, events and data analysis, provides security configuration information to the PLMN from external applications, and converts information between the PLMN and external parties.

[0158] 3. NRF 132 is a control plane function provided by the operator, which can be used to maintain real-time information on network functions and services in the network.

[0159] 4. PCF 133 is a control plane function provided by the operator. It mainly supports providing a unified policy framework to control network behavior, provides policy rules to the control layer network functions, and is also responsible for obtaining user subscription information related to policy decisions.

[0160] 5. UDM 134 is a control plane function provided by the operator, responsible for storing information such as the subscriber permanent identifier (SUPI), the generic public subscription identifier (GPSI), and the credential of the subscriber in the operator's network.

[0161] 6. UDR 135 is a control plane function provided by the operator, which provides UDM with the ability to save and retrieve subscription data, PCF with the ability to save and retrieve policy data, and save and retrieve user NF group ID information, etc.

[0162] 7. NWDAF 136 is a control plane function provided by the operator, possessing functions such as data collection, model training, data analysis, and model inference. Among them, NWDAF network elements containing the analytics logical function (AnLF) can be used to infer and derive analytical information and publish analytical services. Analysis can refer to the generation or provision of statistical information and / or predictions based on requests from analytics consumers. NWDAF network elements containing the model training logical function (MTLF) can be used to train machine learning (ML) models or artificial intelligence (AI) models and publish new training services, such as providing trained AI or ML models to the AnLF. For the process of acquiring model-related data, the AnLF can act as a data producer network element; the MTLF can act as a data consumer network element.

[0163] Currently, AnLF can request models from MTLF via the MLModelProvision_Subscribe service or messages. These models can be trained by MTLF using model-related data (e.g., samples). Furthermore, AnLF can specify a dataset using a dataset tag, which can contain model-related data. Therefore, a dataset tag can be used to label a dataset, i.e., to label model-related data. A dataset tag can also be understood as an index of the dataset or the data itself. MTLF can act as a consumer network element for model-related data, acquiring data stored in a data storage network element. The data producer network element can be AnLF, and the data storage network element can be an analytics data repository functional (ADRF) or other network elements used for storing data. In this application, related data can be understood as data used for longitudinal federated learning, such as input data, training data, inference data, model parameters, or sample data.

[0164] 8. AF 141 is a control plane function provided by the operator. It mainly provides corresponding services by interacting with other NFs in the PLMN, such as providing network selection information for roaming UE visits, routing data flows, and accessing NEF131.

[0165] 9. AUSF 137 is a control plane function provided by the operator, usually used for Level 1 authentication, that is, authentication between terminal device 110 (subscribed user) and operator network.

[0166] 10. AMF 138 is a control plane network function provided by the operator network, which is responsible for access control and mobility management of terminal equipment 110 accessing the operator network. This includes functions such as mobility state management, allocation of temporary user identity identifiers, authentication and authorization of users.

[0167] 11. SMF 139 is a control plane network function provided by the operator network. It is responsible for managing the protocol data unit (PDU) sessions of terminal equipment 110 (including session establishment, modification and release), and is used for the selection and reselection of user plane function network elements, Internet Protocol (IP) address allocation of terminal equipment, QoS control, etc.

[0168] It is understandable that the aforementioned network elements or functions can be physical entities in hardware devices, software instances running on dedicated hardware, or virtualization functions instantiated on a shared platform (e.g., a cloud platform). Simply put, an NF can be implemented in hardware or software.

[0169] In Figure 1, Nnef, Nnrf, Npcf, Nudm, Nudr, Nnwdaf, Naf, Nausf, Namf, Nsmf, N1, N2, N3, N4, and N6 are interface sequence numbers. For example, the meaning of the above interface sequence numbers can be found in the definitions in the 3GPP standard protocol; this application does not limit the meaning of the above interface sequence numbers. It should be noted that the interface names between the various network functions in Figure 1 are merely examples, and the interface names in this system architecture may be other names; this application does not limit them. Furthermore, the names of the messages (or signaling) transmitted between the above network elements are also merely examples and do not constitute any limitation on the function of the messages themselves.

[0170] It should be noted that in the architecture shown in Figure 1, the interface between (R)AN and CN can also be called the NG interface (not shown in the figure). (R)AN and CN are connected through the NG interface. The NG interface can include the NG-C interface and the NG-U interface. The NG-C interface is the control plane interface, connecting (R)AN and AMF, and is used to transmit control plane data. The NG-U interface is the user plane interface, connecting (R)AN and UPF, and is used to transmit user plane data.

[0171] It should be understood that the network architecture 100 described above is only a network architecture described from the perspective of a service-based architecture. In this service-based architecture, the PLMN can combine some or all network functions on demand and in an orderly manner according to specific scenario requirements, thereby customizing network capabilities and services, and deploying dedicated networks for different services, i.e., realizing 5G network slicing. Network slicing technology enables operators to respond to customer needs more flexibly and quickly, and supports the flexible allocation of network resources.

[0172] It should also be understood that the above naming is defined solely for the purpose of distinguishing different functions and should not constitute any limitation on this application. This application does not preclude the possibility of using other naming conventions in 5G networks and other future networks. For example, in 6G networks, some or all of the above-mentioned network elements may use the terminology from 5G, or may use other names, etc.

[0173] Figure 2 is a schematic diagram of a network architecture 200. As shown in Figure 2, in this Artificial Intelligence / Machine Learning (AI / ML) architecture, an AI / ML enabler client (or AI / ML enablement client) is deployed on the UE. The AI / ML enabler client has AI / ML functions, specifically including federated learning local model training and / or model inference. An AI / ML enabler server (or AI / ML enablement server) is deployed on the network side. The AI / ML enabler server has AI / ML functions, specifically including AI / ML member selection and / or allocating federated learning tasks and assigning tasks to AI / ML enabler clients. Optionally, a central server, AI / ML centralserver, can also be deployed in this network architecture. This AI / ML centralserver can be used to store trained AI / ML models. For example, the AI / ML centralserver is connected to the AI / ML enablement server, facilitating the AI / ML enablement server to report trained AI / ML models to the AI / ML centralserver.

[0174] As an example, the AI / ML enablement server requests at least one AI / ML enablement client to perform model training. Each client trains the model locally, obtains the training results, and reports them to the AI / ML enablement server. The server then aggregates the training results from each client to obtain a trained AI / ML model, which is then reported to the AI / ML central server for storage. Optionally, this trained AI / ML model can be used for applications such as network performance analysis, network anomaly detection, or face recognition.

[0175] Optionally, this application does not limit the specific names of the network elements involved in the network architecture described in Figure 2.

[0176] For ease of understanding, the following describes the relevant terms, concepts, or technologies that may be involved in the embodiments of this application:

[0177] 1. Federated learning;

[0178] Federated learning (FL) is a distributed machine learning method where multiple participants interact with model parameters through secure mechanisms without exchanging or sharing raw training data, thereby achieving collaborative training effects. In other words, federated learning is an encrypted distributed machine learning technology. Federated learning fully utilizes the data and computing power of participating parties, enabling multiple parties to collaboratively build general and robust machine learning models without sharing data. Therefore, federated learning can effectively help multiple organizations use data and perform learning modeling while meeting user privacy protection, data security, and regulatory requirements. Essentially, federated learning aims to share knowledge and parameters without exchanging any of their own data.

[0179] Federated learning includes horizontal federated learning and vertical federated learning. In an environment of increasingly stringent data regulations, federated learning can address key issues such as data ownership, data privacy, data access rights, and access to heterogeneous data. In a data matrix, each horizontal row represents a training sample, and each vertical column represents a data feature. Horizontal federated learning combines multiple rows of samples with the same features from multiple participants, meaning the training data for each participant is horizontally partitioned. Horizontal federated learning is also called feature-aligned federated learning, where the data features of the participants are aligned. Horizontal federated learning can increase the total number of training samples. FL, as a machine learning technique, can be used to solve model training and inference problems when participants are unwilling to share their original data. It is suitable for situations where there is significant overlap in participant training sample identifications (IDs) but minimal overlap in participant data features. FL combines different data features from common samples from multiple participants, meaning the training data for each participant is vertically partitioned; hence, it is called vertical federated learning. Vertical federated learning combines different data features from common samples from multiple participants, meaning the training data for each participant is vertically partitioned. Vertical federated learning, also known as sample-aligned federated learning, means that the training samples of the participants are aligned. Vertical federated learning can increase the feature dimension of the training data.

[0180] 2. Analysis identifier (analysis ID or analytics ID);

[0181] An analysis identifier can be used to indicate an analysis business or service, or simply a service. This service is associated with a model, meaning the model can be used to execute the service; or, in other words, the analysis identifier is associated with a model, meaning the model is used to execute the service corresponding to the analysis identifier.

[0182] Alternatively, it can be understood that MTLF is associated with an analytics identifier, meaning that the model provided by the MTLF supports the execution of the service corresponding to that analytics identifier. For example, an MTLF can be associated with one or more analytics identifiers. This can be understood as the MTLF providing a model for the service corresponding to each of the one or more analytics identifiers. For instance, if MTLF1 is associated with analytics identifier 1 and analytics identifier 2, meaning MTLF1 corresponds to analytics identifier 1 and analytics identifier 2, then MTLF1 can provide a model for the service corresponding to analytics identifier 1, and a model for the service corresponding to analytics identifier 2.

[0183] The terminology used in this application has been briefly explained above, and will not be repeated in the embodiments below. Furthermore, the above explanations of terminology are for ease of understanding only and do not limit the scope of protection of the embodiments of this application.

[0184] Currently, the AI / ML enabler server can initiate a federated learning process, meaning that the AI / ML enabler server and AI / ML enabler client can perform federated learning training. The specific process of federated learning training can be seen in the relevant description in Figure 3 below. For ease of description and understanding, Figure 3 uses the AI / ML enabler server as the server NWDAF and the AI / ML enabler client as the client NWDAF as an example. Here, NRF and NF can be considered as network elements in the 3GPP network system shown in Figure 2.

[0185] Figure 3 is a flowchart of a federated learning execution method. As shown in Figure 3, it includes the following steps. For parts not described in detail, please refer to existing protocols.

[0186] S301, the consumer sends a subscription request message #1 to the server NWDAF (e.g., FL server NWDAF).

[0187] Correspondingly, the server NWDAF receives subscription request message #1 from the consumer.

[0188] The subscription request message #1 is used to subscribe to ML model provisioning or training. For example, a consumer (e.g., an NWDAF containing AnLF, or an NWDAF containing MTLF) uses the Nnwdaf_MLModelProvision service to send a subscription request message #1, such as "subscription request for ML model provisioning / training," to the server NWDAF to retrieve an ML model.

[0189] The subscription request message #1 includes one or more of the following: analytics ID, ML model metric (e.g., ML model accuracy), accuracy reporting interval, and scheduled status (ML model accuracy threshold or time required for the ML model). It should be understood that the ML model accuracy threshold can be used to indicate the target ML model accuracy during training; when the ML model accuracy threshold is reached during training, the server NWDAF can stop the training process. If the consumer provides a time required for the ML model, the server NWDAF can consider this information to determine the maximum response time for its client NWDAF.

[0190] S302, the server NWDAF determines the client NWDAF(s).

[0191] For example, the server NWDAF can perform the client NWDAF discovery process with the NRF, and the specific implementation can be found in existing related descriptions. For instance, the client NWDAF(s) registers with the NRF, the server NWDAF sends a discovery request message (Nnrf_NFDiscovery_Request) to the NRF to request the discovery of at least one client NWDAF performing the federated learning task, the NRF authorizes, for example, based on the type of the desired NF / NF service and the type of the server NWDAF, whether to allow the server NWDAF to discover the desired NF instance, and the NRF sends a discovery response message (Nnrf_NFDiscovery_Response) to the server NWDAF, which includes information about the client NWDAF(s) determined by the server NWDAF. The client NWDAF(s) can be one or more, such as client NWDAF1, ..., client NWDAF N (e.g., FL client NWDAF), without limitation.

[0192] S303, the server NWDAF sends a subscription request message #2 to the client NWDAF(s).

[0193] Correspondingly, the client NWDAF(s) receives subscription request message #2 from the server NWDAF.

[0194] For example, the subscription request message #2 could be Nnwdaf_MLModelTraining_Subscribe, such as when the server NWDAF sends Nnwdaf_MLModelTraining_Subscribe or Nnwdaf_MLModelTrainingInfo_Request to the client NWDAF(s) to request the execution of local model training.

[0195] The subscription request message #2 carries one or more of the following: initial federated learning parameters provisioning, ML model metrics, initial ML model, or maximum response time, where maximum response time refers to the maximum response time for the client NWDAF to report temporary local ML model information to the server NWDAF.

[0196] S304, optionally, the client NWDAF(s) collects data.

[0197] For example, if a client NWDAF does not already have local data available, each client NWDAF can collect its local data from the NF (data provider) using the current mechanism in TS23.288.

[0198] S305, the client NWDAF(s) sends a subscription response message #2 to the server NWDAF.

[0199] Correspondingly, the server NWDAF receives subscription response message #2 from the client NWDAF(s).

[0200] For example, the subscription response message #2 could be Nnwdaf_MLModelTraining_Notify, used to report training information of the local model. For instance, during federated learning training, each client NWDAF trains an initial ML model provided by the server NWDAF based on its own data and reports temporary local ML model information to the server NWDAF in Nnwdaf_MLModelTraining_Notify or Nnwdaf_MLModelTrainingInfo_Response.

[0201] Optionally, Nnwdaf_MLModelTraining_Notify or Nnwdaf_MLModelTrainingInfo_Response may also include local ML model metrics and training input data information computed by the client NWDAF(s) (e.g., the region covered by the dataset, the sampling ratio, the maximum / minimum value of each dimension of the data, etc.).

[0202] Optionally, the ML model sent from the client NWDAF(s) to the server NWDAF during federated learning training is the information required by the server NWDAF to build an aggregate model based on the locally trained ML model.

[0203] Optionally, if the client NWDAF cannot complete the training of the temporary local ML model within the maximum response time provided by the server NWDAF, the client NWDAF(s) can send a delay notification, including a delay event indication, an optional cause code (e.g., local ML model training failed, or more time is required for local ML model training), and the expected time available for the client NWDAF to complete training before the maximum response time has elapsed. In other words, the client NWDAF(s) can report a temporary local model, or the server NWDAF can notify the client NWDAF(s) to skip this round of training.

[0204] S306, Optionally, the server NWDAF sends a subscription request message #3 to the client NWDAF(s).

[0205] Correspondingly, the client NWDAF(s) receives subscription request message #3 from the server NWDAF.

[0206] For example, the subscription response message #3 could be `Nnwdaf_MLModelTraining_Notify`, carrying the extended response time and / or the current iteration round ID. For instance, if the server NWDAF receives a notification / response from the client NWDAF(s) that it cannot complete training within the maximum response time, the server NWDAF can send an extended maximum response time `_MLModel TrainingInfo_Request` to the client NWDAF in `Nnwdaf_MLModelTraining_Subscribe` or `Nnwdaf`. Before this request, the client NWDAF needs to report temporary local ML model information to the server NWDAF. Otherwise, the server NWDAF might instruct the client NWDAF to skip reporting for this iteration. The client NWDAF includes the current iteration round ID in the request message to indicate that this request is used to modify the training parameters for the current iteration round.

[0207] Optionally, the server NWDAF can notify the client NWDAF to stop ML model training by sending a termination request and report the current local ML model update.

[0208] S307, the client-side NWDAF performs model aggregation.

[0209] For example, the client-side NWDAF aggregates all local ML model information retrieved in step S305 to update the global ML model. Optionally, the server-side NWDAF may also compute a global ML model metric, such as based on local ML model metrics, or by applying the global model to the validation dataset (if available). The server-side NWDAF may update the global ML model each time the client-side NWDAF provides updated local ML model information, or the server-side NWDAF may decide to wait for local ML model information from all client-side NWDAFs before updating the global ML model.

[0210] If the server NWDAF provides a maximum response time for the client NWDAF to provide temporary local ML model information in step S303, or provides an extended maximum response time in step S306, the server NWDAF decides to wait for client NWDAFs that have not yet provided their temporary local ML model within the (extended) maximum response time, or simply aggregate the retrieved local ML model information instances to update the global ML model. The server NWDAF makes its decision based on the client NWDAF's notification / response, or if no notification is received, based on its local configuration.

[0211] S308, optionally, the client NWDAF sends an update message to the consumer.

[0212] Correspondingly, the consumer receives update messages from the client NWDAF.

[0213] The update message is used to indicate the current ML training status to the consumer. For example, in response to step S301, the client NWDAF sends an Nnwdaf_MLModelProvision_Notify message to the consumer to periodically (e.g., a certain number of training rounds or every 10 minutes) or dynamically update the global ML model metrics (e.g., reaching an ML model accuracy threshold or training time expiring) to the consumer when certain predetermined states are reached.

[0214] S309, Optionally, the consumer sends a subscription request message #4 to the server NWDAF.

[0215] Correspondingly, the server NWDAF receives subscription request message #4 from the consumer. This subscription request message #4 is used to modify the subscription to update or terminate it.

[0216] For example, the consumer decides whether the current model meets their requirements, such as global ML model metrics indicating consumer satisfaction, and decides to stop or continue the training process. The user can re-invoke the Nnwdaf_MLModelProvision_Subscribe service operation used in step S301 to stop or continue the training process.

[0217] S310, optionally, the server NWDAF updates or terminates the federated learning training process.

[0218] For example, based on the subscription request message #4 sent by the consumer in step S309, the server NWDAF updates or terminates the current federated learning training process. Optionally, if the server NWDAF receives a request to stop the federated training process in step S309, steps S311 and S312 are skipped.

[0219] S311, Optionally, the server NWDAF sends aggregate model information to the client NWDAF(s).

[0220] Correspondingly, the client NWDAF(s) receives aggregated model information from the server NWDAF.

[0221] In other words, if the federated learning training process continues, the server NWDAF will determine the client NWDAF and send an Nnwdaf_MLModelTraingInfo_Request containing aggregated ML model information to the selected client NWDAF(s) for the next round of federated training.

[0222] S312, Optionally, the client NWDAF(s) updates the local model based on the aggregated model information.

[0223] For example, each client NWDAF updates its local ML model based on the aggregated ML model information distributed by the server NWDAF in step S311.

[0224] It should be noted that by repeating the above steps S304-S312, the training termination condition is reached (e.g., the maximum number of iterations, or the result of the loss function is below a threshold).

[0225] When the federated training process is complete, the server NWDAF requests the client NWDAF(s) to terminate the federated learning training process. In one implementation, the server NWDAF calls the Nnwdaf_MLModelTraing_Unsubscribe service with the reason code that the federated learning process has been completed and the final aggregated ML model information is available. The client NWDAF(s) then terminates its local model training. If it has received the final aggregated ML model information from the server NWDAF, the client NWDAF(s) can store the aggregated ML model information for further use.

[0226] Based on the federated learning training process shown in Figure 3, there may be situations where client NWDAFs (e.g., one or more of client NWDAF1, ..., client NWDAF N) are unable to complete temporary local ML model training within the maximum response time provided by the server NWDAF. For ease of description, client NWDAFs that cannot complete temporary local ML model training can be collectively referred to as source AI / ML members. In this case, the source AI / ML can request that the intermediate training results or intermediate model information obtained from local training be passed to other AI / ML members, so that other AI / ML members can continue to perform model training tasks. For ease of description, federated learning members that obtain intermediate training results or intermediate model information from source AI / ML can be referred to as target AI / ML members.

[0227] Figure 4 is a flowchart of a method for transmitting intermediate models. As shown in Figure 4, it includes the following steps. For parts not described in detail, please refer to existing protocols.

[0228] S401, the source AI / ML member sends a request message to the AI / ML enablerserver, and correspondingly, the AI / ML enablerserver receives the request message from the source AI / ML member.

[0229] The request message carries requirement information, including at least one of the following: operation type (ML model training), intermediate model information (e.g., MLmodel information, which may include model status or operation results), data transmission requirements (e.g., data volume, data transmission completion time, or data transmission required resources), or, instruction information, to instruct the AI / MLenablerserver to perform the transmission.

[0230] For example, when a source AI / ML member needs to transmit intermediate model information, the source AI / ML member sends a request message carrying the required information to the AI / MLenablerserver, requesting the AI / MLenablerserver to perform the transmission.

[0231] S402, AI / MLenablerserver discovers target AI / ML member.

[0232] For example, the AI / MLenablerserver discovers target AI / ML members based on the requested information. The specific discovery process can be found in existing related descriptions and will not be explained here.

[0233] S403, the AI / MLenabler server sends a request message to the target AI / ML member, and the target AI / ML member receives the request message from the AI / MLenabler server.

[0234] S404, Target AI / ML member authentication and authorization, please refer to existing relevant descriptions for details, which will not be explained here.

[0235] S405, the target AI / ML member sends a response message to the AI / MLenablerserver, and correspondingly, the AI / MLenablerserver receives the response message from the target AI / ML member.

[0236] This response message is used to instruct the target AI / ML member to support receiving intermediate model information from the source AI / ML member.

[0237] S406, AI / MLenablerserver determines the transmission mode.

[0238] For example, the transmission modes include: relaying through the AI / MLenablerserver, that is, the AI / MLenablerserver can send intermediate model information from the source AI / ML member to the target AI / ML member; or, the source AI / ML member and the target AI / ML member can directly transmit intermediate model information.

[0239] S407, the AI / MLenablerserver generates AI / ML auxiliary information. For example, transmission time window information, used to indicate the time or period for transmitting intermediate model information.

[0240] S408, intermediate model information is transferred between the source AI / ML member and the target AI / ML member.

[0241] Method 1: The AI / MLenablerserver sends the target AI / ML member's information to the source AI / ML member; the source AI / MLmodel sends the intermediate model information directly to the target AI / ML member.

[0242] Method 2: The source AI / ML member sends intermediate model information to the AI / MLenabler server, and then the AI / MLenabler server forwards the intermediate model information to the target AI / ML member. This can be done through pass-through and is not limited.

[0243] Federated learning (FL), a machine learning technique in the core network, allows multiple participants (e.g., server NWDAF and client NWDAF) to train locally and exchange model parameters through secure mechanisms without interacting with training data, thus achieving collaborative training. For example, during federated learning, a client NWDAF can train a local model based on local data and send the local model information to the server NWDAF. The server NWDAF aggregates the local models obtained from multiple client NWDAFs and then distributes the aggregated model information to each participating client NWDAF, continuing to train the model based on local data. This iterative process continues until the federated learning is complete and a global model is obtained. At this point, the server NWDAF can send the global model generated by federated learning to the client NWDAF, which can then store the global model information for further use.

[0244] However, if the client-side NWDAF cannot complete training, skipping the current training session or reporting a temporary local model may impact overall model performance. Furthermore, if the client-side NWDAF cannot continue training the FL model due to resource or performance issues, and the server-side NWDAF does not select a new client-side NWDAF for training, it may lead to increased convergence time and decreased performance. Moreover, when the source AI / ML member (or the source client-side NWDAF) cannot complete the training task, selecting a target AI / ML member (or the target client-side NWDAF) based solely on data transmission requirements for the current intermediate model, without considering the current training task, may result in the selection of an unsuitable target AI / ML member, preventing it from completing training.

[0245] In view of this, this application provides a communication method and a communication device. The federated learning server determines the second federated learning member based on the capability information of the candidate federated learning members and the first training requirements. That is, when selecting a suitable federated learning member to continue the training task, the current training situation or training requirements are taken into account, which can avoid the increase of model convergence time and ensure model performance.

[0246] The communication method provided by the embodiments of this application will be described in detail below with reference to the accompanying drawings, and can be applied to any of the communication systems shown in Figures 1 to 3 above. It should be understood that the embodiments of this application can be applied to scenarios where the sending end and the receiving end communicate.

[0247] It should also be understood that the embodiments shown below do not specifically limit the structure of the execution subject of the method provided in the embodiments of this application, as long as it is possible to communicate according to the method provided in the embodiments of this application by running the code or program that records the method provided in the embodiments of this application. For example, the method provided in the embodiments of this application can be executed by a federated learning server, a first federated learning member, or a second federated learning member. Unless otherwise specified, "federated learning server, first federated learning member, or second federated learning member" in this application can refer to the device itself, or a component in the device (e.g., a communication module, processor, circuit, chip (such as a modem chip, also known as a baseband chip, or a system-on-a-chip (SoC) or system-in-a-package (SIP) chip containing a modem core, or a chip system, etc.), or it can be a logic module or software that can implement all or part of the device functions.

[0248] Figure 5 is a flowchart illustrating a communication method provided in an embodiment of this application. As shown in Figure 5, the method includes the following steps, illustrated using a federated learning server, a first federated learning member, and a second federated learning member as the executing entities. For parts not described in detail, please refer to the relevant descriptions of existing protocols.

[0249] S510, the first federated learning member sends the first request message to the federated learning server;

[0250] Correspondingly, the federated learning server receives the first request message from the federated learning server.

[0251] Optionally, the first request message can be an AI / ML memberdiscoveryrequest message, and the specific message name is not limited.

[0252] In this application, the federated learning server can be an AI / ML server, an AI / ML enabling server (AI / MLenablerserver or AI / ML E server), or a server NWDAF, i.e., a Server NWDAF containing MTLF, an NWDAF with federated learning server capabilities. For example, the federated learning server supports at least one of the following: discovering federated learning members (FL members), AI / ML enabling clients (AI / MLenablerclient or AI / ML E client), AI / ML enabling servers (enablerserver), or client devices (e.g., client NWDAF); requesting client devices to perform local model training and report local model information, specifically, sending a base model or initial model to the client device, thereby requesting the client device to train the base model or initial model using local data to obtain a local model, and reporting local model information; aggregating local model information from client devices and generating a global model; aggregating model information after training the base model from client devices and generating a global model; and returning the global model to the client device to achieve iterative training.

[0253] In this application, the first federated learning member can be an AI / ML client, an AI / ML enabling client (AI / MLenablerclient or AI / ML E client), a client NWDAF, an AI / ML enabling client (enablerclient), or an AI / ML enabling server (enablerserver). The first federated learning member supports executing the training task of the first model. This can be understood as the first federated learning member (or entity, network element, device, or node, etc.) possessing the ability to execute the training task of the first model.

[0254] Optionally, the first federated learning member can be replaced by a first federated learning client or a first client device. The first client device can be a client NWDAF, i.e., a Client NWDAF containing MTLF, an NWDAF with federated learning client capabilities. For example, the first federated learning member supports at least one of the following: generating local model information by using available local data for federated learning training; specifically, receiving information about the base model or initial model sent by the federated learning server and training the base model or initial model using available local data to generate local model information; reporting the trained local model information to the federated learning server; reporting the trained base model information to the federated learning server; receiving global model information from the federated learning server and repeating the training iterations.

[0255] For example, the training task of the first model may include at least one of the following: network performance analysis, AI model training, network service quality analysis, network abnormal behavior detection, face recognition, or cross-institutional medical data analysis and disease prediction.

[0256] Alternatively, federated learning can be replaced with federated machine learning, joint learning, or consortium learning.

[0257] Alternatively, the training task of the first model can be replaced with other names such as federated learning task, federated learning, federated learning function, federated learning process, or federated learning activity.

[0258] Optionally, the number of members in the first federated learning group can be one or more, without limitation.

[0259] In the first implementation, the first request message includes the first training requirement corresponding to the first model.

[0260] The first training requirement refers to the current training requirement for the training task of the first model, or in other words, the first training requirement is determined after the first federated learning member performs the training task of the first model.

[0261] Optionally, before performing step S510, the method further includes the following step S501, which is not shown in the figure.

[0262] S501, the first federated learning member determines the first training requirement based on the intermediate training results of the first model and the second training requirement corresponding to the first model.

[0263] The intermediate training results are obtained by the first federated learning member performing the training task of the first model. For the specific implementation method, please refer to the relevant description of step S502 below. The second training requirement is obtained from the federated learning server. For the specific implementation method, please refer to the relevant description of step 1 of step S502 below.

[0264] It is understood that the first model in this embodiment is requested by the model requester. The federated learning server can determine at least one first federated learning member to participate in the training of the first model. For at least one first federated learning member, there is at least one local model. For example, at least one first federated learning member includes FL member#1 and FL member#2. FL member#1 trains locally to obtain local model #1 according to the second training requirements. This local model #1 is the intermediate training result #1 of the first model. Furthermore, FL member#1 can determine the first training requirement #1 based on the intermediate training result #1 and the second training requirement. Similarly, FL member#2 trains locally to obtain local model #2 according to the second training requirement. This local model #2 is the intermediate training result #2 of the first model. Furthermore, FL member#2 can determine the first training requirement #2 based on the intermediate training result #2 and the second training requirement.

[0265] As an example, if it is determined that the first federated learning member cannot continue training the first model, and / or that the intermediate training results of the first model do not meet the second training requirement, the first federated learning member determines the first training requirement based on the intermediate training results and the second training requirement. For example, the first federated learning member can determine that it cannot continue training based on its current local training situation, the second training requirement, and its available resources or available time for training.

[0266] Understandably, the second training requirement refers to the overall training requirements for the training task of the first model, which can be understood as the training requirements determined before the training task of the first model begins.

[0267] In the second implementation, the first request message includes the intermediate training results of the first model.

[0268] Optionally, before performing step S510 or S501, the method further includes the following step S502, which is not shown in the figure.

[0269] S502, the first federated learning member obtains the intermediate training results of the first model, which may include the following steps.

[0270] Step 1: The model requester sends a third request message to the federated learning server, and the federated learning server receives the third request message from the model requester. The third request message requests the execution of the training task for the first model and includes the second training requirements corresponding to the first model.

[0271] Step 2: The federated learning server determines the first federated learning member from the candidate federated learning members based on their ability information and the second training requirements.

[0272] Step 3: The federated learning server sends a fourth request message to the first federated learning member, and correspondingly, the first federated learning member receives the fourth request message from the federated learning server. The fourth request message requests the execution of the training task for the first model and includes a second training requirement.

[0273] Step 4: The first federated learning member performs the training task of the first model according to the second training requirements and obtains the intermediate training results.

[0274] Optionally, the number of candidate federated learning members can be one or more, without limitation. For example, candidate federated learning members may include a first federated learning member and a second federated learning member.

[0275] Optionally, the third or fourth request message can be an AI / ML modeltrainingrequest message, and the specific message name is not limited.

[0276] Optionally, the model requester can be an application (APP), a client (e.g., NWDAF), a server (e.g., NWDAF), or a third party (e.g., NF, AF, or AS).

[0277] Based on step S502 above, the first federated learning member may send a first request message carrying the intermediate training results to the federated learning server, i.e., execute step S510, if it is determined that the first federated learning member cannot continue to execute the training task of the first model, and / or it is determined that the intermediate training results do not meet the second training requirements.

[0278] Optionally, based on the second implementation method described above, the method further includes the following step S520.

[0279] In the S520, the federated learning server determines the first training requirement based on the intermediate training results.

[0280] As an example, the federated learning server determines the first training requirement based on the intermediate training results of the first model and the second training requirement. For the specific implementation method, please refer to the relevant description of the first federated learning member determining the first training requirement in step S501 above. For the sake of brevity, it will not be described here.

[0281] In summary, the first request message sent by the first federated learning member to the federated learning server can carry either the first training requirement or intermediate training results; this application does not impose any limitation on this. The difference lies in the following: in the first implementation, the federated learning server can directly obtain the first training requirement, while in the second implementation, the federated learning server needs to determine the first training requirement based on the obtained intermediate training results and the second training requirement.

[0282] Optionally, the first request message in step S510 may further include indication information, which indicates that the first federated learning member cannot continue to perform the training task of the first model, and / or, the indication information indicates that at least one federated learning member has been found. For example, the indication information can be indicated by 1 bit, where bit "1" indicates that the first federated learning member cannot continue to perform the training task of the first model, bit "0" indicates that the first federated learning member supports continuing to perform the training task of the first model, and vice versa.

[0283] It is understood that if the first request message does not carry any indication information, it can implicitly indicate that the first federated learning member cannot continue to perform the training task of the first model. That is, the federated learning server can determine that the first federated learning member cannot continue to perform the training task of the first model based on the intermediate training results or the first training requirements carried in the first request message. Alternatively, if the indication information carried in the first request message indicates that the first federated learning member cannot continue to perform the training task of the first model, it can implicitly indicate that the federated learning server has found at least one federated learning member. Or, if the indication information carried in the first request message indicates that at least one federated learning member has been found, it can implicitly indicate that the first federated learning member cannot continue to perform the training task of the first model. This application does not limit this.

[0284] Optionally, the first request message in step S510 above may further include at least one of the following: an identifier of the first model, an identifier of the training task of the first model, an analysis identifier corresponding to the first model, a service identifier (service ID) corresponding to the first model, or an application identifier (App ID) corresponding to the first model, used to indicate the training task of the first model.

[0285] Below, examples are provided to illustrate the intermediate training results of the first model mentioned above, as well as the specific content of the first and second training requirements corresponding to the first model.

[0286] For example, the intermediate training result includes at least one of the following: a temporary model obtained by performing the training task of the first model; the accuracy information of the temporary model; the accuracy information required by the training task of the first model; the training time of the temporary model; the training time required to continue performing the training task of the first model; the resource usage of the temporary model; the training resources required to continue performing the training task of the first model; the number of training rounds of the temporary model; or, the number of training rounds required to continue performing the training task of the first model, as specifically explained below.

[0287] (1) The temporary model obtained by performing the training task of the first model, for example, the temporary model can be a part of the first model that meets the second training requirements;

[0288] (2) Precision information for temporary models indicates the precision or accuracy of a temporary model. It is an indicator of a machine learning model's ability to make predictions or decisions based on data and can be used to compare the performance of different models or evaluate the effectiveness of a specific model for a given task. For example, mean square error (MSE) and mean absolute error (MAE) are commonly used to measure the accuracy of regression models, which are used to predict continuous values. MSE is the average of the squared differences between predicted and true values, while MAE is the average of the absolute differences between predicted and true values. For example, the precision or accuracy of a temporary model is 80%.

[0289] (3) The accuracy information required for the training task of the first model, such as 100%;

[0290] (4) Training time for temporary models, such as 40 minutes, or 10:00:00-10:40:00;

[0291] (5) The training time required to continue the training task of the first model, for example, 20 minutes;

[0292] (6) Resource usage of the temporary model, for example, the size of computing power resources or graphics processing unit (GPU) resources used is 25MB;

[0293] (7) Continue to execute the training task of the first model and the required training resources, such as 25MB of GPU resources;

[0294] (8) The number of training rounds for the temporary model, for example, 15 rounds;

[0295] (9) The number of training rounds required to continue the training task of the first model, for example, 5 rounds.

[0296] For example, the first training requirement includes at least one of the following: the accuracy information required to continue the training task of the first model; the training time required to continue the training task of the first model; the training resources required to continue the training task of the first model; or, the number of training epochs required to continue the training task of the first model, as specifically defined below.

[0297] (1) Continue to perform the training task of the first model to obtain the required accuracy information, such as 100%;

[0298] (2) The training time required to continue the training task of the first model, for example, 20 minutes;

[0299] (3) Continue to execute the training task of the first model and the required training resources, such as 25MB of GPU resources;

[0300] (4) The number of training rounds required to continue the training task of the first model, for example, 5 rounds.

[0301] For example, the second training requirement includes at least one of the following: the analysis identifier corresponding to the first model; the service identifier corresponding to the first model; the application identifier corresponding to the first model; the time for acquiring the first model; the training duration of the first model; the accuracy information required by the training task of the first model; the training time required by the training task of the first model; the training resources required by the training task of the first model; or, the number of training rounds required by the training task of the first model, as specifically explained below.

[0302] (1) The analysis identifier corresponding to the first model is used to indicate the training task of the first model.

[0303] (2) The service identifier corresponding to the first model is used to indicate the training task of the first model;

[0304] (3) The application identifier corresponding to the first model is used to indicate the training task of the first model;

[0305] For example, the analysis identifier, service identifier, or application identifier corresponding to the first model can be used to indicate a specific function or service that is related to the first model. That is, the first model can be used to execute the specific function or service, or in other words, the first model supports the execution of the specific function or service corresponding to the analysis identifier. This specific function or service can be, for example, facial recognition or network performance analysis. For instance, the analysis identifier can indicate services such as terminal anomaly detection or terminal session analysis (e.g., Quality of Service (QoS) analysis).

[0306] The service name can be Nnwdaf_MLModelTraining service, and the specific message name is not limited.

[0307] (4) Obtain the time of the first model, for example, the time of obtaining the first model is required to be before 10:00:00;

[0308] (5) The training time of the first model, for example, the training time of the first model is required to be 100 minutes;

[0309] (6) The accuracy information required for the training task of the first model, such as 100%;

[0310] (7) The training time required for the training task of the first model, for example, 120 minutes, or 10:00:00-12:00:00;

[0311] (8) The training resources required for the training task of the first model, such as requiring 50MB of GPU resources;

[0312] (9) The number of training rounds required for the training task of the first model, for example, 20 rounds.

[0313] Understandably, the first training requirement can be determined based on the second training requirement and the current training status (e.g., the intermediate training results of the first model). For example, the training time required to continue training the first model can be determined based on the training time required by the training task of the first model and the training time of the temporary model (or intermediate training results) obtained from the training task of the first model; as another example, the number of training epochs required to continue training the first model is determined based on the number of training epochs required by the training task of the first model and the number of training epochs of the temporary model (or intermediate training results) obtained from the training task of the first model.

[0314] As an example, if the second training requirements initially obtained by the first federated learning member include: the training time required for the training task of the first model is 60 minutes and the required GPU resource size is 50MB, and the intermediate training results obtained by the first federated learning member after executing the training task according to the second training requirements include: a temporary model that has been trained locally for 30 minutes and the model training is 50% complete, then if the first federated learning member cannot continue to execute the training task of the first model, the first training requirements that can be determined include: based on the temporary model obtained from local training, the training time required to continue to execute the training task of the first model is 30 minutes and the required GPU resource size is 25MB.

[0315] It should be noted that if the first federated learning member determines that its unavailable time is less than or equal to the remaining time of the overall model training, that is, for the training task of the first model, the second training requirements include: a total of 10 model training and iterations are required, totaling 600 minutes, with each local model training session lasting approximately 60 minutes. If the first federated learning member has already performed 210 minutes of local model training (i.e., the fourth local model training session lasted 30 minutes), and the first federated learning member is unavailable for 30 minutes, then the first federated learning member can instruct the second federated learning member to perform one round of local model training. In this case, the first training requirements for the second federated learning member include: training time of 30 minutes, GPU resource size of 25MB, and 1 round of training task execution.

[0316] S530, the federated learning server determines the second federated learning member from the candidate federated learning members based on the ability information of the candidate federated learning members and the first training requirements.

[0317] The capability information indicates whether the candidate federated learning members support continuing the training task of the first model. At least one of the candidate federated learning members supports continuing the training task of the first model, and at least one federated learning member includes a second federated learning member.

[0318] In this application, the second federated learning member can be an AI / ML client, an AI / ML enabling client (AI / MLenablerclient or AI / MLE client), an AI / ML enabling server (enablerserver), or a client device (e.g., client NWDAF). The second federated learning member supports continuing the training task of the first model. Supporting continuing the training task of the first model can be understood as supporting breakpoint resumption, that is, supporting the ability to continue training the first model after obtaining incomplete intermediate training results.

[0319] Optionally, the number of members in the second federated learning group can be one or more, without limitation.

[0320] Optionally, before performing step S530, the method further includes the following step S503, which is not shown in the figure.

[0321] S503, the federated learning server obtains the capability information of candidate federated learning members.

[0322] As an example, candidate federated learning members send their capability information to the federated learning server. This capability information indicates whether the candidate federated learning member supports resuming training after interruption, that is, whether it supports continuing model training after obtaining incomplete intermediate training results. For example, if the candidate federated learning members include a first federated learning member and a second federated learning member, the first federated learning member can send its capability information to the federated learning server to indicate that it does not support resuming training after interruption; the second federated learning member can send its capability information to the federated learning server to indicate that it supports resuming training after interruption.

[0323] Optionally, in this embodiment, after receiving a third request message from the model requester, the federated learning server can determine to execute the training task of the first model. It can then query the candidate federated learning members to see if they support endpoint continuation training of the first model; that is, to query whether the candidate federated learning members can / are / support continuing the training task of the first model after obtaining intermediate training results. For example, if the candidate federated learning members include a first federated learning member and a second federated learning member, the first federated learning member can report to the federated learning server that it does not support continuing the training task of the first model after obtaining intermediate training results, while the second federated learning member can report to the federated learning server that it supports continuing the training task of the first model after obtaining intermediate training results.

[0324] Optionally, the capability information of the candidate federated learning members can be reported to the learning server by the candidate federated learning members when initiating the registration process, that is, the capability information of the candidate federated learning members can be carried in the registration request message. Alternatively, the capability information of the candidate federated learning members can also be reported independently, that is, the capability information of the candidate federated learning members can be not carried in the registration request message. This application does not limit this.

[0325] In this application, the capability information of the candidate federated learning member is also used to indicate at least one of the following: the training time supported by the candidate federated learning member; the available training time corresponding to the candidate federated learning member; the training resources supported by the candidate federated learning member; the available training resources corresponding to the candidate federated learning member; or, the number of model training epochs supported by the candidate federated learning member, as specifically interpreted below.

[0326] (1) The training time supported by the candidate federated learning members, for example, the training time supported by the first federated learning member is 60 minutes, or 10:00:00-11:00:00, and the training time supported by the second federated learning member is 100 minutes, or 10:00:00-11:40:00.

[0327] (2) The available training time corresponding to the candidate federated learning members, for example, the available training time corresponding to the first federated learning member is 10:30:00-11:00:00, and the available training time corresponding to the second federated learning member is 10:20:00-11:20:00.

[0328] (3) Training resources supported by candidate federated learning members, for example, the first federated learning member supports training resources with a GPU size of 25MB and the second federated learning member supports training resources with a GPU size of 30MB.

[0329] (4) Available training resources corresponding to candidate federated learning members, for example, the available training resources GPU size corresponding to the first federated learning member is 10MB, and the available training resources GPU size corresponding to the second federated learning member is 20MB.

[0330] (5) The number of training rounds of the model supported by the candidate federated learning members, for example, the first federated learning member supports 20 training rounds of the model, and the second federated learning member supports 30 training rounds of the model.

[0331] The above is an example provided for ease of description and understanding, using the example of candidate federated learning members including the first and second federated learning members. Other schemes are not excluded. Optionally, candidate federated learning members may also include a third federated learning member.

[0332] Based on the second federated learning member determined in step S530 above, the first federated learning member or the federated learning server can request the second federated learning member to continue the training task of the first model to ensure the completion of the training of the first model. The following examples, combining methods one and two, illustrate how the first federated learning member and the federated learning server can request the second federated learning member to continue the training task of the first model.

[0333] Method 1: The first federated learning member requests the second federated learning member to continue the training task of the first model.

[0334] S540, the federated learning server sends the first response message to the first federated learning member;

[0335] Correspondingly, the first federated learning member receives the first response message from the federated learning server.

[0336] The first response message includes information about the second federated learning member.

[0337] Optionally, the first response message can be an AI / ML memberdiscoveryresponse message, and the specific message name is not limited.

[0338] For example, the information of the second federated learning member includes at least one of the following: the identity (ID) of the second federated learning member, the NF instance ID of the second federated learning member (e.g., NF Instance Id(s) of the NF producer consumer), the NF instance ID of a network element that has federated learning client functionality or supports federated learning client functionality (e.g., NF Instance Id(s) of the NWDAF), the address information of the second federated learning member, or other information that can identify the second federated learning member. For example, the information of the second federated learning member can be at least one of the following: FL member ID, AI / ML client ID, AI / ML enabler server ID or AI / ML Eserver ID, AI / ML enablerclient ID or AI / ML E client ID, Instance Id(s) of the AI / ML enabler server, or Instance Id(s) of the AI / ML enablerclient, etc.

[0339] Optionally, the first response message may also include at least one of the following: an analysis identifier corresponding to the first model, a service identifier corresponding to the first model, or an application identifier corresponding to the first model, used to indicate the training task of the first model.

[0340] Optionally, the first response message may also include instruction information to instruct the second federated learning member to continue performing the training task.

[0341] S550, the first federated learning member sends a second request message to the second federated learning member;

[0342] Correspondingly, the second federated learning member receives a second request message from the first federated learning member.

[0343] The second request message is used to request the second federated learning member to continue the training task of the first model.

[0344] Optionally, the second request message can be an FL modetransferrequest message, and the specific message name is not limited.

[0345] In one implementation, the second request message includes intermediate training results of the first model and a first training requirement. For example, the intermediate training results indicate that the training time for the local model (or temporary model) trained by the first federated learning member is 40 minutes, such as 10:00:00-10:40:00; the size of the training resources used to train this local model is 25MB; and the number of training epochs for local training is 15. The first training requirement indicates that the training time for continuing the first model should be 30 minutes, such as 11:00:00-11:30:00; the size of the training resources required for continuing the training task of the first model is 15MB; and the number of training epochs required for continuing the training task of the first model is 10. This implementation can implicitly indicate the second training requirement, that is, the model requester's request for the total training time, training resources, and number of training epochs required to execute the training task of the first model.

[0346] In another implementation, the second request message includes intermediate training results of the first model and second training requirements. For example, the intermediate training results indicate that the training time for the local model (or temporary model) trained by the first federated learning member is 40 minutes, e.g., 10:00:00-10:40:00; the size of the training resources used to train this local model is 25MB; and the number of training epochs is 15. The second training requirements indicate that the total training time required to execute the first model is 70 minutes, e.g., it can be completed between 10:00:00 and 12:00:00; the total size of the training resources required to execute the training task of the first model is 40MB; and the total number of training epochs required to execute the training task of the first model is 25. This implementation can implicitly indicate the first training requirements, i.e., the requirements for training time, training resources, and the number of training epochs required for the second federated learning member to continue executing the training task of the first model.

[0347] In another implementation, the second request message includes intermediate training results of the first model, a first training requirement, and a second training requirement. For example, the intermediate training results indicate that the training time for the local model (or temporary model) trained by the first federated learning member is 40 minutes, for example, 10:00:00-10:40:00; the size of the training resources used to train this local model is 25MB; and the number of training epochs for local training is 15. The first training requirement indicates that the training time for continuing the first model should be 30 minutes, for example, 11:00:00-11:30:00; the size of the training resources required for continuing the training task of the first model is 15MB; and the number of training epochs required for continuing the training task of the first model is 10. The second training requirement indicates that the total training time required for executing the first model is 70 minutes, for example, it can be completed within 10:00:00-12:00:00; the total size of the training resources required for executing the training task of the first model is 40MB; and the total number of training epochs required for executing the training task of the first model is 25.

[0348] The above implementation is merely an example for ease of understanding; other solutions are not excluded.

[0349] Optionally, the second request message may also carry information about the federated learning server, such as the identifier and / or address information of the federated learning server, so that the second federated learning member can subsequently feed back the training results obtained from local training to the federated learning server.

[0350] Method 2: The federated learning server requests the second federated learning member to continue the training task of the first model.

[0351] S560, the federated learning server sends a second request message to the second federated learning member;

[0352] Correspondingly, the second federated learning member receives a second request message from the federated learning server.

[0353] The second request message is used to request the second federated learning member to continue the training task of the first model. The second request message includes intermediate training results and the first training requirements.

[0354] Optionally, the second request message can be an ML modetrainingrequest message, and the specific message name is not limited.

[0355] Understandably, the first method described above can be executed based on the first implementation of step S510. That is, when the first federated learning server carries the first training requirement in the first request message, or does not carry the intermediate training results of the first model, the federated learning server can execute the discovery process of the second federated learning server (see the relevant description in Figure 3 for the specific discovery process), and feed back the information of the second federated learning member to the first federated learning member through method one, so that the first federated learning member can request the second federated learning member to continue the training task of the first model. Similarly, the second method described above can be executed based on the second implementation of step S510. That is, when the first federated learning server can carry the intermediate training results of the first model in the first request message, after determining the first training requirement and executing the discovery process of the second federated learning server, the federated learning server can use method two to request the second federated learning member to continue the training task of the first model.

[0356] Based on method one or method two above, the second federated learning member can continue to execute the training task of the first model according to the intermediate training results and the first training requirements, and feed back the training results to the federated learning server. See steps S570-S580 below for details.

[0357] S570, the second federated learning member continues to perform the training task based on the intermediate training results and the first training requirements, and obtains the first model and the training information of the first model.

[0358] For example, the training information of the first model includes at least one of the following: the identifier of the first model (e.g., the ID of the first model); the address information of the first model (addressofmodel); the analysis identifier corresponding to the first model; the service identifier corresponding to the first model; the application identifier corresponding to the first model; the time of acquiring the first model; the training duration of the first model; or, the accuracy information of the first model. For details not covered, please refer to the relevant description above.

[0359] Optionally, before executing step S570, the second federated learning member can determine whether to continue the training task, i.e., the method further includes the following step S504, which is not shown in the figure.

[0360] S504, the second federated learning member determines whether to continue the training mission based on the intermediate training results and the first training requirements.

[0361] As an example, the second federated learning member determines whether to continue the training task based on its own capabilities, intermediate training results, and the first training requirements. For instance, the second federated learning member supports continuing the training task of the first model, with an available training time of 120 minutes and available GPU resources of 50 MB. The intermediate training results include: the first federated learning member has trained 50% of the first model, using 20 MB of GPU resources, with a training time of 60 minutes. The second training requirements indicate that the overall training requirements for the first model include: a required training time of 100 minutes and a required GPU resource size of 60 MB. Therefore, it is evident that the second federated learning member has the ability to continue the training task of the first model.

[0362] As another example, the second federated learning member can receive instruction information from the first federated learning member and determine whether to continue the training task based on the instruction information. This instruction information is used to instruct the second federated learning member to continue the training task. This instruction information can be explicitly carried in the first response message in step S540 above, or it can be determined by the first federated learning member based on information about the second federated learning member carried in the first response message; there is no limitation on this.

[0363] Understandably, steps S570-S580 are only executed if it is determined that the second federated learning member will continue to perform the training task, so that the training of the first model can be completed.

[0364] Understandably, since the federated learning server selects or determines a second federated learning member from the candidate federated learning members in step S530 based on the intermediate training results and the second training requirements, to continue the training task of the first model, it can be said that the second federated learning member supports continuing the training task of the first model, or in other words, the second federated learning member has the ability to continue training the first model after obtaining the incomplete intermediate training results. Therefore, the above step S504 can be omitted, because under normal circumstances, after executing step S504, the second federated learning member can be determined to participate in continuing the training task of the first model.

[0365] Optionally, if the second federated learning member is participating in or about to participate in other federated learning (i.e., different from the training task of the first model) before receiving the second request message, thus causing the second federated learning member to be unable to participate in or complete the training task of the first model, then the above step S504 can be executed, and the second federated learning member can determine that it will not participate in continuing to execute the training task of the first model. In this case, the second federated learning member does not execute steps S570-S580, that is, the second federated learning member can reject the message to the federated learning server. Optionally, a reason value can be carried, which can indicate that the second federated learning member is participating in or about to participate in other federated learning, etc.

[0366] S580, the second federated learning member sends a second response message to the federated learning server;

[0367] Correspondingly, the federated learning server receives a second response message from the second federated learning member.

[0368] The second response message includes the first model and its training information, meaning that the second federated learning member can feed back the local training results to the federated learning server, which includes the first model and its training information.

[0369] Optionally, for the first method described above, if the second request message does not carry information about the federated learning server, the second federated learning member can send the first model and its training information to the first federated learning member, who will then send the first model and its training information to the federated learning server. If the second request message carries information about the federated learning server, the second federated learning member can send the first model and its training information to the federated learning server based on the information, without needing to forward it through the first federated learning member, thus saving signaling overhead.

[0370] Optionally, for Method 2 above, the second federated learning member can send the first model and its training information to the federated learning server after completing local training.

[0371] Optionally, the first model and its training information can be carried in the FL model report (e.g., FL localmodereport).

[0372] Optionally, the second federated learning member may send a second response message to the first federated learning member or the federated learning server to indicate that the second federated learning member confirms the continuation of the training task of the first model, or to indicate that the second federated learning member has completed the training of the first model.

[0373] Optionally, the second response message can be an FL modetransferresponse message or an ML modetrainingresponse message; the specific message name is not limited.

[0374] That is, the second federated learning member can send its trained local model (i.e., the first model) to the federated learning server. After obtaining the local models of all the first and second federated learning members, the federated learning server can perform model aggregation to calculate the global model data. Optionally, the federated learning server can further instruct the federated learning members to perform a new round of local model training, which is not limited.

[0375] Based on the above scheme, the federated learning server, based on the first training requirement or intermediate training results of the first model carried in the received first request message, can select at least one second federated learning member from the candidate federated learning members to continue the training task of the first model, ensuring the completion of the training of the first model. That is, when selecting a suitable second federated learning member to continue the training task, the federated learning server considers the current training situation or training requirements, thus avoiding an increase in model convergence time and a decrease in model performance.

[0376] For ease of understanding, the specific processes applicable to the embodiments of this application are described below with reference to Figures 6 and 7 for different scenarios. In the examples below, the federated learning server is AI / MLserver (or replaced by AI / MLenablerserver or AI / ML E server), the first federated learning member is AI / MLclient#1 (or replaced by AI / MLenablerclient#1 or AI / ML E client#1), the second federated learning member is AI / MLclient#2 (or replaced by AI / MLenablerclient#2 or AI / ML E client#2), and the model requester is App. Optionally, other federated learning members may also be included, such as AI / MLclient#3 (or replaced by AI / MLenablerclient#3 or AI / ML E client#3). This application does not limit the number of federated learning members performing the training task of the first model. It is understood that the processes described below are only illustrative examples, and the embodiments of this application are not limited thereto. Content not described in detail below can be referred to the description in method 500, and will not be repeated here.

[0377] It should be noted that the AI / ML E client in this application embodiment can be replaced by UE, wireless access controller (AC), or VAL server. That is, UE, AC, or VAL server can act as federated learning members to perform federated learning tasks.

[0378] It should be noted that the names of the messages transmitted between the execution entities in this application embodiment are merely examples for ease of understanding, and this application does not impose specific limitations on the message names.

[0379] Figure 6 is a flowchart illustrating the communication method provided in an embodiment of this application. As shown in Figure 6, this implementation involves the first federated learning member requesting the second federated learning member to continue the training task of the first model, specifically including the following steps. For parts not described in detail, please refer to the relevant descriptions of the existing protocols described above.

[0380] S601, the App client (e.g., the App client) sends a registration request message to AI / MLE client#1 (i.e., the first federated learning member), and correspondingly, AI / MLE client#1 receives the registration request message from the App client.

[0381] The registration request message carries at least one of the following: service ID, analytics ID, or APP ID, which indicates the first model or the training task of the first model.

[0382] S602, AI / ML E client#1 sends a registration response message to the App client, and correspondingly, the App client receives the registration response message from AI / ML E client#1.

[0383] The registration response message indicates whether the registration was successful or failed.

[0384] S603, AI / ML E client#1 sends a registration request message to AI / ML E server (i.e., federated learning server), and correspondingly, AI / ML E server receives the registration request message from AI / ML E client#1.

[0385] The registration request message carries the service ID, analytics ID, APP ID, or the capability of AI / ML E client#1 (i.e., the capability information of the first federated learning member, including compute capability or storage capability). For a detailed interpretation of the capability information of the first federated learning member, please refer to the relevant descriptions of steps S530 and S503 of method 500 above.

[0386] Furthermore, the AI / ML E server sends a registration response message to AI / ML E client #1, and correspondingly, AI / ML E client #1 receives the registration response message from the AI / ML E server, which is used to indicate whether the registration was successful or failed.

[0387] Similarly, AI / ML E client #2 sends a registration request message to the AI / ML E server to request registration with the AI / ML E server. This registration request message may carry the capabilities of AI / ML E client #2 (i.e., the capability information of the second federated learning member). For a detailed explanation of the capability information of the second federated learning member, please refer to the relevant descriptions of steps S530 and S503 of method 500 above.

[0388] S604, the App (i.e., the model requester) sends an AI / ML model training request message (i.e., the third request message) to the AI / ML E server, and correspondingly, the AI / ML E server receives the AI / ML model training request message from the App.

[0389] The AI / ML model training request message includes a second training request (ML model training request). Optionally, it may also carry at least one of the following: service ID, analytics ID, or APP ID. For a specific interpretation of the second training request, please refer to the relevant description of step S510 of method 500 above.

[0390] Optionally, step S604 above can be replaced by: the APP can also send a request message to AI / ML client#1 and / or AI / ML client#2 to request the execution of the training task of the first model. Further, AI / ML client#1 and / or AI / ML client#2 send an AI / ML model training request message (i.e., a third request message) to the AI / ML server. Correspondingly, the AI / ML server receives the AI / ML model training request message from AI / ML client#1 and / or AI / ML client#2. That is to say, the AI / ML server can directly or indirectly receive the AI / ML model training request message (i.e., the third request message) from the App, and this application does not limit this.

[0391] S605, AI / ML E server identifies AI / ML E client #1.

[0392] For example, the AI / ML E server determines or selects AI / ML E client #1 to perform the training task of the first model based on the second training requirements and the capability information of AI / ML E client #1. Specifically, the AI / ML E server and NRF can exchange information and execute the process of discovering AI / ML E client #1. The specific discovery process can be referred to the relevant description in Figure 3 above, which will not be explained here.

[0393] S606, the AI / ML E server sends an AI / ML model training request message (i.e., the fourth request message) to the AI / ML E client#1, and correspondingly, the AI / ML E client#1 receives the AI / ML model training request message from the AI / ML E server.

[0394] The AI / ML model training request message is used to request AI / ML client #1 to perform the training task of the first model. The AI / ML model training request message includes a second training requirement, and optionally may also carry at least one of the following: service ID, analytics ID, or APP ID.

[0395] S607, if AI / MLE client #1 determines that it cannot continue the training task, and / or determines that the intermediate training results cannot meet the second training requirement, it determines the first training requirement. Specifically, AI / MLE client #1 determines the first training requirement based on the second training requirement and the intermediate training results.

[0396] For example, AI / MLE client#1 performs local model training according to the second training requirements to obtain intermediate training results. For a detailed explanation of the first training requirements and the intermediate training results, please refer to the relevant description of step S510 of method 510 above.

[0397] Understandably, during the training process, after the AI / MLE client has trained its local model, it can interact with the AI / MLE server. That is, AI / MLE client #1 can send its trained local model to the AI / MLE server. After obtaining the local models of all AI / MLE clients, the AI / MLE server can calculate the global model data. Optionally, the AI / MLE server can further instruct the AI / MLE client to perform a new round of local model training.

[0398] For example, when the AI / ML E client detects that the current resources cannot continue model training, and / or the currently trained model cannot meet the required accuracy, the AI / ML E client#1 can determine the first training requirement based on the current FL model training situation, that is, the ability to determine the target FL member, which is the federated learning capability that the AI / ML E client#2 needs to have. For the specific implementation method, please refer to the relevant description of step S501 of the above method 500.

[0399] S608, AI / MLE client #1 sends a discovery request message (i.e., the first request message, such as AI / MLE member discovery request) to AI / MLE server, and correspondingly, AI / MLE server receives the discovery request message from AI / MLE client #1.

[0400] The discovery request message is used to request the discovery of at least one second federated learning member to continue the training task of the first model. The discovery request message carries the first training requirement. Optionally, the discovery request message may also include at least one of a service ID, analytics ID, or app ID.

[0401] S609, the AI / ML E server determines AI / ML E client #2 based on the first training requirements and the capability information of the candidate AI / ML E clients (e.g., including AI / ML E client #1 and AI / ML E client #2), that is, selects a suitable AI / ML E client #2 as a federated learning member to continue to execute the training task of the first model. For the specific implementation method, please refer to the relevant description of step S530 of the above method 500.

[0402] S610, the AI / ML E server sends a discovery response message to AI / ML E client #1, and correspondingly, AI / ML E client #1 receives the discovery response message from the AI / ML E server.

[0403] The discovery response message includes information about AI / ML E client#2, such as the ID or address of AI / ML E client#2. Optionally, the discovery response message may also include at least one of the following: service ID, analytics ID, or app ID.

[0404] S611, AI / ML E client #1 sends an FL model transfer request message (i.e., a second request message, such as an FL model transfer request) to AI / ML E client #2, and AI / ML E client #2 receives the FL model transfer request message from AI / ML E client #1.

[0405] The FL model transmission request message is used to request AI / ML E client#2 to continue the training task of the first model.

[0406] For example, the FL model transfer request message carries intermediate training results and initial training requirements. Optionally, the FL model transfer request message may also include at least one of a service ID, analytics ID, or app ID. Optionally, the FL model transfer request message may also include information about the AI / ML E server, such as the AI / ML E server's ID and / or address.

[0407] Optionally, the FL model transmission request message may also carry indication information to indicate whether AI / ML E client#2 can continue model training. The method for determining this indication information can be found in the relevant description of step S504 of method 500 above.

[0408] S612, AI / ML E client#2 continues to execute the training task based on the intermediate training results and the first training requirements, and obtains the first model and the training information of the first model.

[0409] For a detailed explanation of the training information for the first model, please refer to the relevant description of step S570 of method 500 above.

[0410] Optionally, before executing step S612, AI / ML E client#2 can determine whether to continue training the model based on the intermediate training results and the first training requirements. For specific implementation details, please refer to the relevant description of step S504 in method 500 above.

[0411] S613, AI / ML E client #2 sends an FL model transfer response message to AI / ML E client #1, and correspondingly, AI / ML E client #1 receives the FL model transfer response message from AI / ML E client #2.

[0412] The FL model transmission response message is used to indicate whether AI / ML E client#2 can continue model training. If model training can continue, AI / ML E client#2 executes the above step S612.

[0413] Optionally, this application does not limit the execution order of steps S612 and S613.

[0414] S614, AI / ML E client #2 sends an FL model report (e.g., FL localmodereport) to AI / ML E server, and correspondingly, AI / ML E server receives the FL model report from AI / ML E client #2. The FL model report includes a first model and its training information.

[0415] S615, AI / ML E server performs model aggregation.

[0416] For details on the implementation of steps S614-S615 above, please refer to the relevant description of step S580 in method 500 above.

[0417] Based on the above scheme, AI / ML E client #1 can determine the capabilities of the target AI / ML E client according to the local model training status and the model training requirements of the AI / ML E server, and feed this information back to the AI / ML E server. This allows the AI / ML E server to select a suitable AI / ML E client #2 to continue the training task. Furthermore, AI / ML E client #1 can send intermediate training results and initial training requirements to AI / ML E client #2, requesting that AI / ML E client #2 can continue the model training without skipping or discarding the current training if AI / ML E client #1 is unable to continue the training task. This avoids increasing the model convergence time and preventing the model performance from degrading.

[0418] Figure 7 is a flowchart illustrating the communication method provided in an embodiment of this application. Unlike Figure 6 above, this implementation involves the federated learning server requesting the second federated learning member to continue the training task of the first model. As shown in Figure 7, this includes several steps; for parts not described in detail, please refer to the relevant descriptions of the existing protocols described above.

[0419] S701, the App client (e.g., the App client) sends a registration request message to AI / ML E client#1 (i.e., the first federated learning member), and correspondingly, AI / ML E client#1 receives the registration request message from the App client.

[0420] S702, AI / ML E client#1 sends a registration response message to the App client, and correspondingly, the App client receives the registration response message from AI / ML E client#1.

[0421] S703, AI / ML E client#1 sends a registration request message to AI / ML E server (i.e., federated learning server), and correspondingly, AI / ML E server receives the registration request message from AI / ML E client#1.

[0422] Similarly, AI / ML E client#2 sends a registration request message to the AI / ML E server to request registration with the AI / ML E server.

[0423] S704, the App (i.e., the model requester) sends an AI / ML model training request message (i.e., the third request message) to the AI / ML E server, and correspondingly, the AI / ML E server receives the AI / ML model training request message from the App.

[0424] The AI / ML model training request message includes a second training requirement (ML model training request).

[0425] S705, AI / ML E server identifies AI / ML E client #1.

[0426] For example, the AI / ML E server determines or selects AI / ML E client #1 to perform the training task of the first model based on the second training requirements and the capability information of AI / ML E client #1.

[0427] S706, the AI / ML E server sends an AI / ML model training request message (i.e., the fourth request message) to the AI / ML E client#1, and correspondingly, the AI / ML E client#1 receives the AI / ML model training request message from the AI / ML E server.

[0428] The AI / ML model training request message is used to request AI / ML client #1 to perform the training task of the first model. The AI / ML model training request message includes a second training requirement.

[0429] S707, AI / ML E client #1 determines a first training requirement if it determines that the training task cannot continue and / or that the intermediate training results cannot meet the second training requirement. Specifically, AI / ML E client #1 determines the first training requirement based on the second training requirement and the intermediate training results.

[0430] For example, AI / ML E client#1 performs local model training according to the second training requirements and obtains intermediate training results.

[0431] S708, AI / ML E client #1 sends a discovery request message (i.e., the first request message, such as AI / ML member discovery request) to AI / ML E server, and correspondingly, AI / ML E server receives the discovery request message from AI / ML E client #1.

[0432] The discovery request message is used to request the discovery of at least one second federated learning member to continue the training task of the first model. The discovery request message carries intermediate training results.

[0433] The specific implementation methods of steps S701-S708 above can be found in the relevant descriptions of steps S601-S608 of method 600 above, and will not be repeated here for the sake of brevity. The difference between the two is that the discovery request message of step S608 in Figure 6 carries the first training requirement, while the discovery request message of step S708 in Figure 7 carries the intermediate training results.

[0434] S709, the AI / ML E server determines AI / ML E client #2 based on the first training requirements and the capability information of the candidate AI / ML E clients (e.g., including AI / ML E client #1 and AI / ML E client #2).

[0435] Specifically, the AI / ML E server determines the first training requirement based on the intermediate training results obtained in step S708 and the second training requirement obtained in step S704. Then, based on the first training requirement and the capability information of the candidate AI / ML E clients, it determines AI / ML E client #2, that is, selects a suitable AI / ML E client #2 as a federated learning member to continue executing the training task of the first model. For specific implementation, please refer to the relevant descriptions of steps S520-S530 of the above method 500.

[0436] S710, the AI / ML E server sends a model training request message (i.e., a second request message, such as an ML model training request) to AI / ML E client #2. Correspondingly, AI / ML E client #2 receives the model training request message from AI / ML E client #1.

[0437] The model training request message is used to request AI / ML E client#2 to continue the training task of the first model.

[0438] For example, the model training request message carries intermediate training results and initial training requirements. Optionally, the model training request message may also include at least one of a service ID, analytics ID, or app ID.

[0439] Optionally, the model training request message may also carry indication information to indicate whether AI / ML E client#2 can continue model training. The method for determining this indication information can be found in the relevant description of step S504 of method 500 above.

[0440] S711, AI / ML E client#2 continues to execute the training task based on the intermediate training results and the first training requirements, and obtains the first model and the training information of the first model.

[0441] S712, AI / ML E client #2 sends a model training response message (i.e., a second response message, such as ML model training response) to AI / ML E client #1, and correspondingly, AI / ML E client #1 receives the model training response message from AI / ML E client #2.

[0442] S713, AI / ML E client #2 sends an FL model report (e.g., FL localmodereport) to AI / ML E server, and correspondingly, AI / ML E server receives the FL model report from AI / ML E client #2. The FL model report includes a first model and its training information.

[0443] S714, AI / ML E server performs model aggregation.

[0444] For details on the implementation of steps S711-S714 above, please refer to the relevant descriptions of steps S612-S615 of method 600 above. For the sake of brevity, they will not be repeated here.

[0445] Based on the above scheme, the AI / ML E server can determine the capabilities of the target AI / ML E client based on the local model training status of AI / ML E client#1 and the model training requirements parameters requested by the AI / ML E server. Then, it can select AI / ML E client#2 to continue the training task. Furthermore, the AI / ML E server can send the intermediate training results and the first training requirements to AI / ML E client#2, requesting that AI / ML E client#2 can continue to execute the model training, without skipping or discarding the current training if AI / ML E client#1 is unable to continue the training task. This avoids increasing the model convergence time and preventing the problem of model performance degradation.

[0446] The technical solution of this application is described above in conjunction with Figures 5 to 7. That is, when the first federated learning member determines that it cannot continue to execute the training task of the first model, and / or the intermediate training result of the first model does not meet the second training requirements corresponding to the first model, it interacts with the federated learning server to find a suitable second federated learning member to continue to execute the training task of the first model. Subsequently, the first federated learning member or the federated learning server requests the second federated learning member to continue to execute the training task, so as to ensure the completion of the training of the first model and ensure the performance of the model.

[0447] The technical solution of this application will be described below with reference to Figures 8 and 9. The difference from Figures 5 to 7 is that, when the first federated learning member determines that it cannot continue to execute the training task of the first model, and / or the intermediate training results of the first model do not meet the second training requirements corresponding to the first model, it can avoid information interaction with the federated learning server, thereby improving model training efficiency. This is because when the federated learning server requests the first federated learning member to execute the training task of the first model, it can also feed back the information of the backup federated learning member (i.e., the second federated learning member) to the first federated learning member.

[0448] Figure 8 is a flowchart illustrating a communication method provided in an embodiment of this application. As shown in Figure 8, the method includes the following steps, with a federated learning server, a first federated learning member, a second federated learning member, and a model requester as the execution entities for illustrative purposes. For parts not described in detail, please refer to Figure 5 above or the relevant descriptions of existing protocols.

[0449] S810, the model requesting direction sends a third request message to the federated learning server;

[0450] Correspondingly, the federated learning server receives third-party request messages from the model requester.

[0451] The third request message is used to request the execution of the training task of the first model. The third request message includes the second training requirements corresponding to the first model. For a detailed explanation of the second training requirements, please refer to the relevant description of step S510 of the above method 500.

[0452] S820, the federated learning server determines the first federated learning member and the second federated learning member from the candidate federated learning members based on the ability information of the candidate federated learning members and the second training requirements.

[0453] Understandably, the second federated learning member is an alternative (or standby) federated learning member used to perform the training task of the first model. Specifically, the second federated learning member can be seen as an alternative federated learning member to the first federated learning member. The alternative (or standby) federated learning member can be understood as follows: the second federated learning member can continue to perform the training task of the first model when the first federated learning member is unable to do so, ensuring that the first model can be trained successfully. In other words, the second federated learning member supports the continued execution of the training task of the first model, or in other words, the second federated learning member has the ability to continue the training task of the first model.

[0454] For a detailed explanation of the competency information of candidate federated learning members, please refer to the relevant description in Method 500 above.

[0455] S830, the federated learning server sends a fourth request message to the first federated learning member;

[0456] Correspondingly, the first federated learning member receives the fourth request message from the federated learning server.

[0457] The fourth request message is used to request the execution of the training task of the first model. The fourth request message includes the second training requirements and information about the second federated learning members.

[0458] S840, the first federated learning member performs the training task of the first model according to the second training requirements and obtains the intermediate training results of the first model.

[0459] For a detailed interpretation of the intermediate training results, please refer to the relevant description of Method 500 above.

[0460] The specific implementation of steps S810-S830 above can be found in the description of steps 1 to 4 in step S502 of method 500 above, and will not be repeated here for the sake of brevity. The difference is that in step S820, the federated learning server also determines the backup second federated learning member, and in step S830, it feeds back the information of the second federated learning member to the first federated learning member.

[0461] S850, if it is determined that the first federated learning member cannot continue to perform the training task of the first model, and / or it is determined that the intermediate training results of the first model do not meet the second training requirements, the first federated learning member determines the first training requirements based on the intermediate training results and the second training requirements corresponding to the first model.

[0462] For a detailed explanation of the first training requirement, please refer to the relevant description of Method 500 above.

[0463] For details regarding the specific implementation of the first training requirement determined by the first federated learning member based on the intermediate training results and the second training requirement corresponding to the first model, please refer to the relevant description of step S501 of the above method 500.

[0464] Since the first federated learning member has already obtained the information of the alternative (or standby) federated learning members in step S830, if it is determined that the first federated learning member cannot continue to execute the training task of the first model, and / or the intermediate training results of the first model do not meet the second training requirements, the first federated learning member can directly request the second federated learning member to continue to execute the training task of the first model, that is, execute the following step S860, without interacting with the federated learning server to find alternative federated learning members, which can improve the efficiency of model training.

[0465] S860, the first federated learning member sends a second request message to the second federated learning member;

[0466] Correspondingly, the second federated learning member receives a second request message from the first federated learning member.

[0467] The second request message is used to request the second federated learning member to continue the training task of the first model. The second request message includes the intermediate training results of the first model and the first training requirements.

[0468] S870, the second federated learning member continues to perform the training task based on the intermediate training results and the first training requirements, and obtains the first model and the training information of the first model.

[0469] For a detailed explanation of the training information for the first model, please refer to the relevant description of step S570 of method 500 above.

[0470] Optionally, before executing step S870, the second federated learning member can determine whether to continue the training task. For the specific implementation method, please refer to the relevant description of step S504 in method 500 above.

[0471] S880, the second federated learning member sends the first model and its training information to the federated learning server;

[0472] Correspondingly, the federated learning server receives the first model and its training information from the second federated learning member.

[0473] For the specific implementation of steps S860-S880 above, please refer to the relevant descriptions of steps S550, S570 and S580 of method 500 above. For the sake of brevity, they will not be repeated here.

[0474] Based on the above scheme, the federated learning server, based on the second training requirements carried in the received third request message and the capability information of the candidate federated learning members, can select a first federated learning member and a backup learning member of the first federated learning member (i.e., a second federated learning member) from the candidate federated learning members to continue the training task of the first model, ensuring the completion of the first model's training. That is, if the first federated learning member subsequently determines that it cannot continue the training task of the first model, and / or the intermediate training results of the first model do not meet the corresponding second training requirements, it can directly request the second federated learning member to continue the training task of the first model without interacting with the federated learning server. This improves model training efficiency, avoids increasing model convergence time, and prevents model performance degradation.

[0475] For ease of understanding, the following explanation, with reference to Figure 9, uses an AI / ML E server (or AI / MLenablerserver or AI / ML E server) as the federated learning server, AI / ML E client#1 (or AI / MLenablerclient#1 or AI / ML E client#1) as the first federated learning member, AI / ML E client#2 (or AI / MLenablerclient#2 or AI / ML E client#2) as the second federated learning member, and an App as the model requester. Optionally, other federated learning members may also be included, such as AI / ML E client#3 (or AI / MLenablerclient#3 or AI / ML E client#3). This application does not limit the number of federated learning members performing the training task of the first model. It is understood that the process described below is only an example, and the embodiments of this application are not limited thereto. Content not described in detail below can be referred to the description in method 800, and will not be repeated here.

[0476] It should be noted that the AI / ML E client in the embodiments of this application can be replaced by UE, AC, or VAL server. That is, UE, AC, or VAL server can act as federated learning members to perform federated learning tasks.

[0477] It should be noted that the names of the messages transmitted between the execution entities in this application embodiment are merely examples for ease of understanding, and this application does not impose specific limitations on the message names.

[0478] Figure 9 is a flowchart illustrating the communication method provided in an embodiment of this application. As shown in Figure 9, it includes the following steps. For parts not described in detail, please refer to Figure 8 above or the relevant description of existing protocols.

[0479] S901, the App client (e.g., the App client) sends a registration request message to AI / ML E client#1 (i.e., the first federated learning member), and correspondingly, AI / ML E client#1 receives the registration request message from the App client.

[0480] S902, AI / ML E client#1 sends a registration response message to the App client, and correspondingly, the App client receives the registration response message from AI / ML E client#1.

[0481] S903, AI / ML E client#1 sends a registration request message to AI / ML E server (i.e., federated learning server), and correspondingly, AI / ML E server receives the registration request message from AI / ML E client#1.

[0482] Similarly, AI / ML E client#2 sends a registration request message to the AI / ML E server to request registration with the AI / ML E server.

[0483] S904, the App (i.e., the model requester) sends an AI / ML model training request message (i.e., the third request message) to the AI / ML E server, and correspondingly, the AI / ML E server receives the AI / ML model training request message from the App.

[0484] The AI / ML model training request message includes a second training requirement (ML model training request).

[0485] For the specific implementation of steps S901-S904 above, please refer to the relevant description of steps S601-S604 in method 600 above. For the sake of brevity, it will not be repeated here.

[0486] S905, the AI / ML E server identifies AI / ML E client #1 and AI / ML E client #2.

[0487] For specific implementation details, please refer to the relevant description of step S820 in method 800 above.

[0488] In this context, AI / ML E client#2 can be seen as a candidate federated learning member of AI / ML E client#1. That is, AI / ML E client#2 can continue to execute the training task of the first model when AI / ML E client#1 is unable to continue to execute the training task of the first model, so as to ensure that the first model can be trained. In other words, AI / ML E client#2 supports continuing to execute the training task of the first model, or in other words, AI / ML E client#2 has the ability to continue to execute the training task of the first model.

[0489] For example, the AI / ML E server determines, based on the second training requirements and the capability information of the candidate AI / ML E clients (e.g., at least AI / ML E client#1 and AI / ML E client#2), whether AI / ML E client#1 will perform the training task of the first model, and selects AI / ML E client#2 as a candidate federated learning member of AI / ML E client#1.

[0490] S906, the AI / ML E server sends an AI / ML model training request message (i.e., the fourth request message) to the AI / ML E client#1, and correspondingly, the AI / ML E client#1 receives the AI / ML model training request message from the AI / ML E server.

[0491] The AI / ML model training request message is used to request AI / ML E client #1 to perform the training task of the first model. The AI / ML model training request message includes a second training requirement and information about AI / ML E client #2.

[0492] S907, if AI / ML E client #1 determines that it cannot continue executing the training task, and / or determines that the intermediate training results cannot meet the second training requirement, it determines the first training requirement. Specifically, AI / ML E client #1 determines the first training requirement based on the second training requirement and the intermediate training results. For the specific implementation method, refer to the relevant description of step S850 of method 800 above.

[0493] For example, AI / ML E client#1 performs local model training according to the second training requirements and obtains intermediate training results.

[0494] S908, AI / ML E client #1 sends a model training request message (i.e., a second request message, such as an ML model training request) or an FL model transfer request message (i.e., a second request message, such as an FL model transfer request) to AI / ML E client #2. Correspondingly, AI / ML E client #2 receives the model training request message or the FL model transfer request message from AI / ML E client #1.

[0495] The model training request message or the FL model transfer request message is used to request AI / ML E client#2 to continue the training task of the first model. For example, the model training request message or the FL model transfer request message carries intermediate training results and the first training requirement.

[0496] S909, AI / ML E client#2 continues to execute the training task based on the intermediate training results and the first training requirements, and obtains the first model and the training information of the first model.

[0497] Optionally, before performing step S909, AI / ML E client#2 can determine whether to continue training the model based on the intermediate training results and the first training requirement.

[0498] S910, AI / ML E client #2 sends a model training response message (i.e., a second response message, such as ML model training response) or an FL model transfer response message (i.e., a second response message, such as FL model transfer response) to AI / ML E client #1. Correspondingly, AI / ML E client #1 receives the model training response message or the FL model transfer response message from AI / ML E client #2.

[0499] S911, AI / ML E client #2 sends an FL model report (e.g., FL localmodereport) to AI / ML E server, and correspondingly, AI / ML E server receives the FL model report from AI / ML E client #2. The FL model report includes a first model and its training information.

[0500] S912, AI / ML E server performs model aggregation.

[0501] For details on the implementation of steps S908-S912 above, please refer to the relevant descriptions of steps S611-S615 in method 600 above.

[0502] Based on the above scheme, the AI / ML E server, based on the second training requirements carried in the received third request message and the capability information of the candidate AI / ML E clients, can select AI / ML E client#1 and its backup learning member, namely AI / ML E client#2, from the candidate AI / ML E clients to continue executing the training task of the first model, ensuring the completion of the first model's training. That is, if AI / ML E client#1 subsequently determines that it cannot continue executing the training task of the first model, and / or the intermediate training results of the first model do not meet the second training requirements corresponding to the first model, it can directly request AI / ML E client#2 to continue executing the training task of the first model without interacting with the AI / ML E server. This improves model training efficiency, avoids increasing model convergence time, and prevents model performance degradation.

[0503] It should be understood that the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0504] It should also be understood that this application will present various aspects, embodiments, or features in relation to systems that may include multiple devices, components, modules, etc. It should be understood and appreciated that individual systems may include additional devices, components, modules, etc., and / or may not include all the devices, components, modules, etc. discussed in conjunction with the accompanying drawings. Furthermore, combinations of these approaches are also possible.

[0505] It should also be understood that in some of the above embodiments, the examples are mainly based on devices in existing network architectures (it should be understood that the specific form of the device is not limited in the embodiments of this application. For example, any device that can achieve the same function in the future is applicable to the embodiments of this application).

[0506] It is understood that the methods and operations implemented by the device (e.g., the second federated learning member, the federated learning server, the first federated learning member) in the above-described method embodiments can also be implemented by components of the device (e.g., chips or circuits).

[0507] The communication method provided in the embodiments of this application has been described in detail above with reference to Figures 1 to 9. The above communication method is mainly described from the perspective of the interaction between the second federated learning member, the federated learning server, and the first federated learning member. It is understood that, in order to realize the above functions, the second federated learning member, the federated learning server, and the first federated learning member include the corresponding hardware structure and / or software module for executing each function.

[0508] Those skilled in the art will recognize that, based on the units and algorithm steps described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is implemented in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0509] The communication device provided in the embodiments of this application will be described in detail below with reference to Figures 10 and 11. The description of the device embodiments corresponds to the description of the method embodiments. Therefore, for contents not described in detail, please refer to the method embodiments above. For the sake of brevity, some contents are not described.

[0510] This application embodiment can divide the communication device into functional modules according to the above method example. For example, each function can be divided into its own functional modules, or two or more functions can be integrated into one processing module. The integrated modules can be implemented in hardware, software, or a combination of both. The module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation. The following description uses the division of functional modules according to each function as an example.

[0511] Figure 10 is a schematic block diagram of a communication device 1000 provided in an embodiment of this application. As shown in Figure 10, the communication device 1000 includes a processing module 1010 and a communication module 1020. The communication device 1000 can be a transmitting device, or a communication device applied to or used in conjunction with a transmitting device to implement a method executed by the transmitting device, such as a chip, chip system, or circuit; or, the communication device 1000 can be a receiving device, or a communication device applied to or used in conjunction with a receiving device to implement a method executed by the receiving device, such as a chip, chip system, or circuit.

[0512] The communication module can also be called a transceiver module, transceiver, transceiver unit, or transceiver device. The processing module can also be called a processor, processing board, processing unit, or processing device. Optionally, the communication module is used to execute the sending and receiving operations of the sending and receiving devices in the above method. The device in the communication module that implements the receiving function can be considered a receiving unit, and the device in the communication module that implements the sending function can be considered a sending unit; that is, the communication module includes a receiving unit and a sending unit.

[0513] Optionally, the communication device 1000 may also include a storage module 1030 for storing device program code and / or data.

[0514] In one example, when the communication device 1000 is applied to the federated learning server, the processing module 1010 can be used to implement the processing function of the federated learning server in the above embodiments, and the communication module 1020 can be used to implement the sending and receiving function of the federated learning server in the above embodiments.

[0515] In another example, when the communication device 1000 is applied to the first federated learning member, the processing module 1010 can be used to implement the processing function of the first federated learning member in the above embodiment, and the communication module 1020 can be used to implement the sending and receiving function of the first federated learning member in the above embodiment.

[0516] In another example, when the communication device 1000 is applied to the second federated learning member, the processing module 1010 can be used to implement the processing function of the second federated learning member in the above embodiments, and the communication module 1020 can be used to implement the sending and receiving function of the second federated learning member in the above embodiments.

[0517] Furthermore, it should be noted that the aforementioned communication module and / or processing module can be implemented through virtual modules. For example, the processing module can be implemented through software functional units or virtual devices, and the communication module can be implemented through software functions or virtual devices. Alternatively, the processing module or communication module can also be implemented through physical devices, such as chips / circuits (e.g., integrated circuits or logic circuits). The communication module can be an input / output circuit and / or a communication interface, performing input operations (corresponding to the aforementioned receiving operation) and output operations (corresponding to the aforementioned sending operation); the processing module is an integrated processor, microprocessor, or circuit (e.g., integrated circuits or logic circuits).

[0518] The module division in this application is illustrative and represents only one logical functional division. In actual implementation, other division methods are possible. Furthermore, the functional modules in the various examples of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0519] In one example, the functional unit in any of the above devices may be one or more integrated circuits configured to implement the above methods, such as: one or more application-specific integrated circuits (ASICs), or one or more central processing units (CPUs), one or more microcontroller units (MCUs), one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0520] In one example, storage module 1030 may include random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory and / or registers, etc.

[0521] Figure 11 is a schematic block diagram of a communication device 2000 provided in an embodiment of this application. Optionally, the communication device 2000 may be a chip or a chip system. Optionally, in this application, the chip system may be composed of chips or may include chips and other discrete devices.

[0522] As shown in Figure 11, the communication device 2000 can be used to implement the functions of any device (e.g., terminal device, network device) in the communication system described in the foregoing examples. The communication device 2000 may include at least one processor 2010. Optionally, the processor 2010 is coupled to a memory, which may be located within the device, integrated with the processor, or located outside the device. For example, the communication device 2000 may also include at least one memory 2020. The memory 2020 stores the computer programs, computer programs or instructions, and / or data necessary for implementing any of the above examples; the processor 2010 may execute the computer programs stored in the memory 2020 to complete the methods in any of the above examples.

[0523] The communication device 2000 may also include a communication interface 2030, through which the communication device 2000 can interact with other devices. Exemplarily, the communication interface 2030 may be a transceiver, circuit, bus, module, pin, or other type of communication interface. When the communication device 2000 is a chip-based device or circuit, the communication interface 2030 in the device 2000 may also be an input / output circuit, capable of inputting information (or receiving information) and outputting information (or sending information). The processor 2010 may be an integrated processor, microprocessor, integrated circuit, or logic circuit, etc., and the processor can determine the output information based on the input information.

[0524] In one example, when the communication device 2000 is applied to the federated learning server, the processor 2010 can be used to implement the processing function of the federated learning server in the above embodiments, and the communication interface 2030 can be used to implement the sending and receiving function of the federated learning server in the above embodiments.

[0525] In another example, when the communication device 2000 is applied to the first federated learning member, the processor 2010 can be used to implement the processing function of the first federated learning member in the above embodiments, and the communication interface 2030 can be used to implement the sending and receiving function of the first federated learning member in the above embodiments.

[0526] In another example, when the communication device 2000 is applied to the second federated learning member, the processor 2010 can be used to implement the processing function of the second federated learning member in the above embodiments, and the communication interface 2030 can be used to implement the sending and receiving function of the second federated learning member in the above embodiments.

[0527] The coupling in this application refers to indirect coupling or communication connection between devices, units, or modules, which can be electrical, mechanical, or other forms, used for information exchange between devices, units, or modules. The processor 2010 may operate in conjunction with the memory 2020 and the communication interface 2030. This application does not limit the specific connection medium between the processor 2010, the memory 2020, and the communication interface 2030.

[0528] Optionally, as shown in FIG11, the processor 2010, the memory 2020, and the communication interface 2030 are interconnected via a bus 2040. Optionally, the bus may include buses of the types such as address bus, data bus, and control bus. Furthermore, for ease of illustration, FIG11 shows one bus 2040, but does not indicate that there is only one bus or only one type of bus.

[0529] It should be understood that the processor mentioned in the embodiments of this application can be one of the following devices or a portion of the circuitry used for processing functions: CPU, and can also be other general-purpose processors, DSPs, ASICs, FPGAs, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0530] It should also be understood that the memory mentioned in the embodiments of this application can be volatile memory and / or non-volatile memory. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM). For example, RAM can be used as an external cache. By way of example and not limitation, RAM includes the following forms: static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM).

[0531] It should be noted that when the processor is a general-purpose processor, DSP, ASIC, FPGA, or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, the memory (storage module) can be integrated into the processor.

[0532] It should also be noted that the memory described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0533] This application also provides a computer-readable storage medium storing computer instructions for implementing the methods executed by a communication device (e.g., a federated learning server, a first federated learning member, or a second federated learning member) in the above-described method embodiments.

[0534] This application also provides a computer program product comprising instructions which, when executed by a computer, implement the methods described above, performed by a communication device (e.g., a federated learning server, a first federated learning member, or a second federated learning member).

[0535] This application also provides a communication system, which includes the federated learning server and / or the first federated learning member in the above embodiments.

[0536] Optionally, the communication system may also include a second federated learning member as described in the above embodiments.

[0537] The explanations and beneficial effects of the relevant contents in any of the devices provided above can be referred to the corresponding method embodiments provided above, and will not be repeated here.

[0538] In the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0539] This application will present various aspects, embodiments, or features relating to systems that may include multiple devices, components, modules, etc. It should be understood and appreciated that individual systems may include additional devices, components, modules, etc., and / or may not include all the devices, components, modules, etc. discussed in conjunction with the accompanying drawings. Furthermore, combinations of these approaches are also possible.

[0540] In this application, examples may reference each other without logical contradiction. For example, methods and / or terms between method embodiments may reference each other, functions and / or terms between device embodiments may reference each other, and functions and / or terms between device examples and method examples may reference each other.

[0541] It should be understood that the above embodiments are mainly illustrated using devices in existing network architectures as examples, and the specific form of the devices is not limited in the embodiments of this application. For example, any device that can achieve the same function in the future is applicable to the embodiments of this application.

[0542] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0543] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be described again here.

[0544] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0545] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0546] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0547] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to existing solutions, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, external hard drives, ROM, RAM, magnetic disks, or optical disks.

[0548] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A communication method, characterized in that, Applications to federated learning servers include: Receive a first request message from the first federated learning member, the first request message including a first training requirement corresponding to the first model; Based on the capability information of the candidate federated learning members and the first training requirement, a second federated learning member is determined from the candidate federated learning members. The capability information indicates whether the candidate federated learning member supports continuing to execute the training task of the first model. At least one of the candidate federated learning members supports continuing to execute the training task of the first model, and the at least one federated learning member includes the second federated learning member.

2. A communication method, characterized in that, Applications to federated learning servers include: Receive a first request message from a first federated learning member, the first request message including intermediate training results of the first model, the intermediate training results being obtained by the first federated learning member performing the training task of the first model; The first training requirement corresponding to the first model is determined based on the intermediate training results; Based on the capability information of the candidate federated learning members and the first training requirement, a second federated learning member is determined from the candidate federated learning members. The capability information indicates whether the candidate federated learning member supports continuing to execute the training task of the first model. At least one of the candidate federated learning members supports continuing to execute the training task of the first model, and the at least one federated learning member includes the second federated learning member.

3. The method according to claim 1 or 2, characterized in that, The first request message further includes indication information, which indicates that the first federated learning member cannot continue to perform the training task of the first model, and / or, the indication information indicates that the at least one federated learning member has been found.

4. The method according to any one of claims 1 to 3, characterized in that, The method further includes: Send a first response message to the first federated learning member, the first response message including information about the second federated learning member.

5. The method according to any one of claims 1 to 3, characterized in that, The method further includes: A second request message is sent to the second federated learning member. The second request message is used to request the continuation of the training task of the first model. The second request message includes the first training requirement and the intermediate training results of the first model.

6. The method according to any one of claims 1 to 5, characterized in that, The intermediate training results of the first model include at least one of the following: A temporary model obtained by performing the training task of the first model; The accuracy information of the temporary model; The accuracy information required for the training task of the first model; The training time of the temporary model; The training time required to continue the training task of the first model; Resource usage of the temporary model; The training resources required to continue the training task of the first model; The number of training rounds for the temporary model; or... The number of training rounds required to continue training the first model.

7. The method according to any one of claims 1 to 6, characterized in that, Before determining a second federated learning member from the candidate federated learning members based on their capability information and the first training requirement, the method further includes: Obtain the capability information of the candidate federated learning members.

8. The method according to any one of claims 1 to 7, characterized in that, The capability information is also used to indicate at least one of the following: The training time supported by the candidate federated learning members; The available training time corresponding to the candidate federated learning members; The training resources supported by the candidate federated learning members; Available training resources corresponding to the candidate federated learning members; or, The number of model training rounds supported by the candidate federated learning members.

9. The method according to any one of claims 1 to 8, characterized in that, The first training requirement includes at least one of the following: Continue to execute the accuracy information required for the training task of the first model; The training time required to continue the training task of the first model; The training resources required to continue the training task of the first model; or, The number of training rounds required to continue training the first model.

10. The method according to any one of claims 1 to 9, characterized in that, Before receiving the first request message from the first federated learning member, the method further includes: Receive a third request message from the model requester, the third request message being used to request the execution of the training task of the first model, the third request message including a second training requirement corresponding to the first model; Based on the capability information of the candidate federated learning members and the second training requirements, the first federated learning member is determined from the candidate federated learning members; A fourth request message is sent to the first federated learning member. The fourth request message is used to request the execution of the training task of the first model. The fourth request message includes the second training requirement. The first training requirement is determined based on the second training requirement and the intermediate training results of the first model.

11. The method according to claim 10, characterized in that, The second training requirement includes at least one of the following: The analysis identifier corresponding to the first model; The service identifier corresponding to the first model; The application identifier corresponding to the first model; Obtain the time of the first model; The training time of the first model; The accuracy information required for the training task of the first model; The training time required for the training task of the first model; The training resources required for the training task of the first model; or, The number of training rounds required for the training task of the first model.

12. The method according to any one of claims 1 to 11, characterized in that, The first federated learning member or the second federated learning member is an enabling client or an enabling server.

13. A communication method, characterized in that, Applicable to first federated learning members, including: If it is determined that the first federated learning member cannot continue to perform the training task of the first model, and / or if it is determined that the intermediate training result of the first model does not meet the second training requirement corresponding to the first model, the first training requirement corresponding to the first model is determined according to the intermediate training result and the second training requirement, wherein the intermediate training result is obtained by the first federated learning member performing the training task of the first model; Send a first request message to the federated learning server, the first request message including the first training requirement; Receive a first response message from the federated learning server, the first response message including information about a second federated learning member, the second federated learning member supporting the continuation of the training task of the first model.

14. A communication method, characterized in that, Applicable to first federated learning members, including: If it is determined that the first federated learning member cannot continue to execute the training task of the first model, and / or if it is determined that the intermediate training result of the first model does not meet the second training requirement corresponding to the first model, a first request message is sent to the federated learning server, wherein the first request message includes the intermediate training result, the intermediate training result is used to determine the first training requirement corresponding to the first model, and the intermediate training result is obtained by the first federated learning member executing the training task of the first model. Receive a first response message from the federated learning server, the first response message including information about a second federated learning member, the second federated learning member supporting the continuation of the training task of the first model.

15. The method according to claim 13 or 14, characterized in that, The method further includes: A second request message is sent to the second federated learning member. The second request message is used to request to continue the training task of the first model. The second request message includes the first training requirement and the intermediate training results.

16. The method according to any one of claims 13 to 15, characterized in that, The first request message further includes indication information, which indicates that the first federated learning member cannot continue to perform the training task of the first model, and / or, the indication information indicates that the at least one federated learning member has been found.

17. The method according to any one of claims 13 to 16, characterized in that, Before sending the first request message to the federated learning server, the method further includes: The capability information of the first federated learning member is sent to the federated learning server, the capability information indicating that the first federated learning member supports the training task of the first model.

18. The method according to any one of claims 13 to 17, characterized in that, Before determining the first training requirement corresponding to the first model based on the intermediate training results and the second training requirement, or before sending the first request message to the federated learning server, the method further includes: Receive a fourth request message from the federated learning server, the fourth request message being used to request the execution of the training task of the first model, the fourth request message including the second training requirement; The training task of the first model is performed according to the second training requirements to obtain the intermediate training results.

19. The method according to any one of claims 13 to 18, characterized in that, The first training requirement includes at least one of the following: Continue to execute the accuracy information required for the training task of the first model; The training time required to continue the training task of the first model; The training resources required to continue the training task of the first model; or, The number of training rounds required to continue training the first model.

20. The method according to any one of claims 13 to 19, characterized in that, The intermediate training results include at least one of the following: A temporary model obtained by performing the training task of the first model; The accuracy information of the temporary model; The accuracy information required for the training task of the first model; The training time of the temporary model; The training time required to continue the training task of the first model; Resource usage of the temporary model; The training resources required to continue the training task of the first model; The number of training rounds for the temporary model; or... The number of training rounds required to continue training the first model.

21. A communication method, characterized in that, Applicable to second federated learning members, including: Receive a second request message from the first federated learning member or the federated learning server. The second request message is used to request to continue the training task of the first model. The second request message includes the intermediate training results of the first model and the first training requirements corresponding to the first model. Based on the intermediate training results and the first training requirements, the training task continues to be executed to obtain the first model and the training information of the first model.

22. The method according to claim 21, characterized in that, Before continuing the training task based on the intermediate training results and the first training requirement, the method further includes: Based on the capability information of the second federated learning member and the first training requirement, it is determined to continue the training task.

23. The method according to claim 21 or 22, characterized in that, Before receiving a second request message from the first federated learning member or the federated learning server, the method further includes: The capability information of the second federated learning member is sent to the federated learning server, the capability information indicating that the second federated learning member supports continuing the training task of the first model.

24. The method according to claim 23, characterized in that, The capability information is also used to indicate at least one of the following: the training time supported by the second federated learning member, the available training time corresponding to the second federated learning member, the training resources supported by the second federated learning member, the available training resources corresponding to the second federated learning member, or the number of model training rounds supported by the second federated learning member.

25. The method according to any one of claims 21 to 24, characterized in that, The intermediate training results include at least one of the following: A temporary model obtained by performing the training task of the first model; The accuracy information of the temporary model; The accuracy information required for the training task of the first model; The training time of the temporary model; The training time required to continue the training task of the first model; Resource usage of the temporary model; The training resources required to continue the training task of the first model; The number of training rounds for the temporary model; or... The number of training rounds required to continue training the first model.

26. The method according to any one of claims 21 to 25, characterized in that, The training information of the first model includes at least one of the following: the identifier of the first model, the address information of the first model, the analysis identifier corresponding to the first model, the service identifier corresponding to the first model, the application identifier corresponding to the first model, the time of acquiring the first model, the training duration of the first model, or the accuracy information of the first model.

27. The method according to any one of claims 21 to 26, characterized in that, The first training requirement includes at least one of the following: the accuracy information required to continue the training task of the first model, the training time required to continue the training task of the first model, the training resources required to continue the training task of the first model, or the number of training rounds required to continue the training task of the first model.

28. The method according to any one of claims 21 to 27, characterized in that, The first federated learning member or the second federated learning member is an enabling client or an enabling server.

29. A communication method, characterized in that, Applications to federated learning servers include: Receive a third request message from the model requester, the third request message being used to request the execution of the training task of the first model, the third request message including a second training requirement corresponding to the first model; Based on the capability information of the candidate federated learning members and the second training requirement, a first federated learning member and a second federated learning member are determined from the candidate federated learning members. The capability information indicates whether the candidate federated learning members support continuing to execute the training task of the first model. At least one of the candidate federated learning members supports continuing to execute the training task of the first model. The at least one federated learning member includes the second federated learning member, which is a candidate federated learning member for executing the training task of the first model. A fourth request message is sent to the first federated learning member. The fourth request message is used to request the execution of the training task of the first model. The fourth request message includes the second training requirements corresponding to the first model and the information of the second federated learning member.

30. The method according to claim 29, characterized in that, Before determining the first and second federated learning members from the candidate federated learning members based on their capability information and the second training requirements, the method further includes: Obtain the capability information of the candidate federated learning members.

31. The method according to claim 29 or 30, characterized in that, The capability information is also used to indicate at least one of the following: the training time supported by the candidate federated learning member, the available training time corresponding to the candidate federated learning member, the training resources supported by the candidate federated learning member, the available training resources corresponding to the candidate federated learning member, or the number of model training rounds supported by the candidate federated learning member.

32. The method according to any one of claims 29 to 31, characterized in that, The second training requirement includes at least one of the following: the analysis identifier corresponding to the first model, the service identifier corresponding to the first model, the application identifier corresponding to the first model, the time of acquiring the first model, the training duration of the first model, the accuracy information required by the training task of the first model, the training time required by the training task of the first model, the training resources required by the training task of the first model, or the number of training rounds required by the training task of the first model.

33. The method according to any one of claims 29 to 32, characterized in that, The first federated learning member or the second federated learning member is an enabling client or an enabling server.

34. A communication method, characterized in that, Applicable to first federated learning members, including: Obtain information about the second federated learning member, who supports continuing the training task of the first model; If it is determined that the first federated learning member cannot continue to perform the training task of the first model, and / or if it is determined that the intermediate training result of the first model does not meet the second training requirement corresponding to the first model, the first training requirement corresponding to the first model is determined according to the intermediate training result and the second training requirement. A second request message is sent to the second federated learning member. The second request message is used to request the continuation of the training task. The second request message includes the first training requirement and the intermediate training results.

35. The method according to claim 34, characterized in that, The process of obtaining information about the second federated learning member includes: Receive a fourth request message from the federated learning server, the fourth request message being used to request the execution of the training task of the first model, the fourth request message including the second training requirements and information of the second federated learning member; The method further includes: The training task is performed according to the second training requirement to obtain the intermediate training results.

36. The method according to claim 34 or 35, characterized in that, Before obtaining information about the second federated learning member, the method further includes: The capability information of the first federated learning member is sent to the federated learning server, the capability information indicating that the first federated learning member supports the training task of the first model.

37. The method according to claim 36, characterized in that, The capability information is also used to indicate at least one of the following: the training time supported by the first federated learning member, the available training time corresponding to the first federated learning member, the training resources supported by the first federated learning member, the available training resources corresponding to the first federated learning member, or the number of model training rounds supported by the first federated learning member.

38. The method according to any one of claims 34 to 37, characterized in that, The first training requirement includes at least one of the following: the accuracy information required to continue the training task of the first model, the training time required to continue the training task of the first model, the training resources required to continue the training task of the first model, or the number of training rounds required to continue the training task of the first model.

39. The method according to any one of claims 34 to 38, characterized in that, The second training requirement includes at least one of the following: the analysis identifier corresponding to the first model, the service identifier corresponding to the first model, the application identifier corresponding to the first model, the time of acquiring the first model, the training duration of the first model, the accuracy information required by the training task of the first model, the training time required by the training task of the first model, the training resources required by the training task of the first model, or the number of training rounds required by the training task of the first model.

40. The method according to any one of claims 34 to 39, characterized in that, The intermediate training results include at least one of the following: a temporary model obtained by performing the training task of the first model, the accuracy information of the temporary model, the accuracy information required by the training task of the first model, the training time of the temporary model, the training time required to continue performing the training task of the first model, the resource usage of the temporary model, the training resources required to continue performing the training task of the first model, the number of training rounds of the temporary model, or the number of training rounds required to continue performing the training task of the first model.

41. The method according to any one of claims 34 to 40, characterized in that, Based on the intermediate training results and the second training requirements, the first training requirements corresponding to the first model are determined, including: Based on the training time required by the training task of the first model and the training time of the temporary model obtained by executing the training task of the first model, determine the training time required to continue executing the training task of the first model; and / or, Based on the number of training rounds required for the training task of the first model and the number of training rounds of the temporary model obtained by executing the training task of the first model, determine the number of training rounds required to continue executing the training task of the first model.

42. The method according to any one of claims 34 to 41, characterized in that, The first federated learning member or the second federated learning member is an enabling client or an enabling server.

43. A communication device, characterized in that, It includes modules for implementing the method as described in any one of claims 1 to 12, or modules for implementing the method as described in any one of claims 13 to 20, or modules for implementing the method as described in any one of claims 21 to 33, or modules for implementing the method as described in any one of claims 34 to 42.

44. A communication device, characterized in that, The method includes at least one processor for executing a computer program or instructions to cause the method as described in any one of claims 1 to 12 to be executed, or to cause the method as described in any one of claims 13 to 20 to be executed, or to cause the method as described in any one of claims 21 to 33 to be executed, or to cause the method as described in any one of claims 34 to 42 to be executed.

45. The communication device according to claim 44, characterized in that, The communication device further includes a memory for storing the computer program or instructions; and / or, The communication device further includes a communication interface coupled to the at least one processor, the communication interface being used for inputting and / or outputting information.

46. ​​A communication system, characterized in that, It includes a federated learning server and / or a first federated learning member, wherein the federated learning server is used to perform the method as described in any one of claims 1 to 12, the first federated learning member is used to perform the method as described in any one of claims 13 to 20, or is used to perform the method as described in any one of claims 34 to 42.

47. The communication system according to claim 46, characterized in that, The communication system further includes a second federated learning member, wherein the second federated learning member is used to perform the method as described in any one of claims 21 to 33.

48. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program that, when run on a computer, causes the method as described in any one of claims 1 to 12 to be performed, or causes the method as described in any one of claims 13 to 20 to be performed, or causes the method as described in any one of claims 21 to 33 to be performed, or causes the method as described in any one of claims 34 to 42 to be performed.

49. A computer program product, characterized in that, Includes a computer program or instructions that, when executed by a processor, cause the method as claimed in any one of claims 1 to 12 to be performed, or cause the method as claimed in any one of claims 13 to 20 to be performed, or cause the method as claimed in any one of claims 21 to 33 to be performed, or cause the method as claimed in any one of claims 34 to 42 to be performed.

Images