Communication method and apparatus

Abstract

The embodiments of the present application relate to the field of communications. Provided are a communication method and apparatus, which can determine a participating entity participating in a VFL task and / or a role of the participating entity in the VFL task. The method comprises: a first apparatus receiving information of first samples corresponding to candidate participating entities which support a first vertical federated learning (VFL) task and serve as a first role, which information is from a second apparatus; and on the basis of the information of the first samples corresponding to the candidate participating entities which serve as the first role, determining a target participating entity of the first VFL task and / or a role of the target participating entity in the first VFL task, so as to send to the second apparatus information of the target participating entity participating in the first VFL task and / or information of the role of the target participating entity in the first VFL task, wherein the first apparatus is an apparatus initiating the first VFL task.

Description

Communication method and apparatus

[0001] The present application claims priority from the Chinese patent application No. 202410445515.6 filed on April 12, 2024, and entitled "Communication method and apparatus", the content of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0002] The present application relates to the field of communication, in particular to a communication method and apparatus. BACKGROUND

[0003] Vertical federated learning (VFL) is a federated learning of different data features of joint samples of multiple participating entities (such as network elements in a vertical federated alliance), that is, the training data of each participating entity is vertically divided. VFL is also called sample-aligned federated learning, that is, the training samples of participating entities are aligned, and VFL can increase the dimension of training data features.

[0004] Currently, the 3rd generation partnership project (3GPP) supports training and / or inference through VFL tasks to implement mobile network related analysis services. Before performing a VFL task, it is necessary to determine the participating entities that finally participate in the VFL task and their roles in the VFL task (such as a master participant or a slave participant). However, in the participating entity determination process defined in the 3GPP, since the role information of the participating entities is introduced, after the VFL task initiator triggers the VFL task, the influence of the role information on the process of how to determine the participating entities participating in the VFL task and / or the roles of the participating entities in the VFL task is not clear. SUMMARY

[0005] The present application provides a communication method and apparatus, which can determine the participating entities participating in a VFL task and / or their roles in the VFL task in the granularity of roles.

[0006] To achieve the above object, the present application adopts the following technical solutions:

[0007] In a first aspect, a communication method is provided. The method can be performed by a first device, by a component of the first device, such as a processor, a chip, or a chip system of the first device, or by a logic module or software that can implement all or part of the first device. The method includes: receiving, by the first device, information about candidate participating entities supporting a first vertical federated learning (VFL) task as first samples corresponding to a first role from a second device, wherein the first device is a device initiating the first VFL task; and determining, by the first device, target participating entities of the first VFL task and / or roles of the target participating entities in the first VFL task according to the information about the candidate participating entities supporting the first VFL task as the first samples corresponding to the first role.

[0008] Based on the communication method, the first device as a task initiator and / or a target participating entity obtains information about samples of candidate participating entities in a first role from the second device as a task coordinator, and the information about the samples in the first role is filtered by the second device using samples supported by the first device. Thus, the candidate participating entities as the target participating entities and / or the roles of the candidate participating entities as the target participating entities in the first VFL task can be determined according to the information about the samples in the first role. Therefore, the target participating entities and / or the roles of the target participating entities can be efficiently determined by using the sample information in the role granularity.

[0009] In the embodiments of the present application, the task initiator can refer to a requester of a VFL task or a requester of a VFL service, and is used to request to initiate the VFL task or the VFL service.

[0010] In a possible design, the method of the first aspect can further include: sending, by the first device, information about the target participating entities participating in the first VFL task and / or information about the roles of the target participating entities in the first VFL task to the second device. Thus, the first device can inform the second device of the determined target participating entities and / or the roles of the target participating entities, so as to facilitate the second device to trigger the target participating entities to perform subsequent model training and / or inference related to the first VFL task.

[0011] In a possible design, the target participating entity can be a primary participant and / or a secondary participant in the first VFL task. The primary participant is configured to provide labels or to provide labels and training data, and the secondary participant is configured to provide training data. For a target participating entity, it can participate in the first VFL task as a primary participant and a secondary participant, or as one of the primary participant and the secondary participant. i i i i i i

[0012] In a possible design, the first sample is an intersection of the second sample and a third sample corresponding to the candidate participating entity in the first role, and the second sample is a sample supported by the first device. In the embodiments of the present application, since the first device is a mandatory participant in the first VFL task, the first device is configured to determine the sample information of the target participating entity and / or the candidate participating entity in the corresponding role, and the sample information of the candidate participating entity in the first role is obtained by filtering the sample that is not supported by the first device from the sample supported by the first device. Therefore, the sample information obtained by the first device can be more accurate, and the efficiency of determining the target participating entity and / or the corresponding role can be improved.

[0013] In a possible design, the method of the first aspect can further include: the first device sending, to the second device, information of a sample of the target participating entity for performing the first VFL task. In the embodiments of the present application, since the second device is a coordinator of the first VFL task, it needs to send the sample data aligned by the sample to each target participating entity as an intermediate node for performing the first VFL task. Therefore, after determining the target participating entity and / or the corresponding role, the first device as the target participating entity can send the information of the sample of the target participating entity for performing the first VFL task to the second device, so as to facilitate subsequent model training and / or inference related to the first VFL task.

[0014] ​​​​​​In a possible design, the target participant entity of the first VFL task and / or the role of the target participant entity in the first VFL task can also be determined according to at least one of the following: a weight of a feature of a sample corresponding to a role supported by the candidate participant entity, a membership weight corresponding to the candidate participant entity, a membership corresponding to the candidate participant entity, or a role supported by the candidate participant entity in the first VFL. In this way, the first device can further determine the target participant entity of the first VFL task and / or the role of the target participant entity based on the sample information of the candidate participant entity at the granularity of the first role, in combination with one or more of the feature weight, the membership requirement, or the role requirement, to more efficiently complete the selection of the target participant entity and / or the role.

[0015] In a second aspect, a communication method is provided. The method can be performed by a second device, or by a component of the second device, for example, a processor, a chip, or a chip system of the second device, and can also be implemented by a logic module or software that can implement all or part of the second device. The method includes: the second device obtaining information of a candidate participant entity supporting a first VFL task as a third sample corresponding to a first role. The second device determines information of the candidate participant entity as a first sample corresponding to the first role based on information of a second sample supported by a first device and information of the candidate participant entity as the third sample corresponding to the first role, the first sample being an intersection of the second sample and the third sample, and the first device being a device initiating the first VFL task. The second device sends the information of the candidate participant entity as the first sample corresponding to the first role to the first device.

[0016] Based on the communication method, the second device as a task coordinator can filter out samples that are not supported by the first device from the information of the candidate participant entity supporting the first VFL task and having the third sample at the granularity of the first role, by using the information of the second sample of the task initiator and the target participant entity, and send the information of the first sample of the candidate participant entity at the granularity of the first role obtained through the filtering to the first device. In this way, the information of the samples that are not supported by the first device among the candidate participant entities can be avoided to be leaked, the privacy and security of the sample data can be ensured, and the first device can efficiently determine the target participant entity and / or the role of the target participant entity based on the information of the first sample of the candidate participant entity at the granularity of the first role obtained through the filtering.

[0017] In a possible design, the method in the second aspect can further include: receiving, by the second device, information of a target participant entity participating in the first VFL task and / or information of a role of the target participant entity in the first VFL task from the first device.

[0018] In a possible design, the information of the target participating entity participating in the first VFL task and / or the information of the role of the target participating entity in the first VFL task is determined according to the information of the first sample corresponding to the candidate participating entity as the first role.

[0019] In a possible design, the role of the target participating entity in the first VFL task is a primary participant or a secondary participant.

[0020] In a possible design, the second device obtaining the information of the third sample corresponding to the candidate participating entity as the first role in supporting the first VFL task can include that the second device sends a first request message to the candidate participating entity, where the first request message is used to request to obtain the information of the sample corresponding to the candidate participating entity as the first role. The second device receives a first response message from the candidate participating entity, where the first response message includes the information of the third sample corresponding to the candidate participating entity as the first role. In this way, the second device can obtain the information of the third sample corresponding to the candidate participating entity as the first role by sending the first request message to the candidate participating entity.

[0021] Specific descriptions of technical effects of the method of the second aspect can be referred to the descriptions of the technical effects of the method of the first aspect, which are not repeated here.

[0022] In a third aspect, a communication method is provided. The method can be executed by a second device, or by a component of the second device, for example, a processor, a chip, or a chip system of the second device, or by a logic module or software that can implement all or part of the second device. The method includes: the second device obtaining information of a third sample corresponding to a candidate participating entity as a first role in supporting a first VFL task. The second device determines a target participating entity of the first VFL task and / or a role of the target participating entity in the first VFL task according to the information of the third sample corresponding to the candidate participating entity as the first role.

[0023] Based on the communication method, the second device as a task coordinator obtains information of samples of each candidate participating entity as a first role in supporting a first VFL task, and determines a candidate participating entity as a target participating entity and / or a role of the candidate participating entity as the target participating entity in the first VFL task according to the information of the samples as the first role.

[0024] In a possible design, the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task can also be determined according to at least one of the following: a weight of a feature of a sample corresponding to a role supported by the candidate participating entity, a membership weight corresponding to the candidate participating entity, a membership corresponding to the candidate participating entity, or a role supported by the candidate participating entity in the first VFL. In this way, the second device as the task coordinator can determine the target participating entity of the first VFL task and / or the role of the target participating entity on the basis of the sample information of the candidate participating entity at the granularity of the first role, and in combination with one or more of the feature weight, the membership requirement, or the role requirement, to more efficiently complete selection of the target participating entity and / or the role.

[0025] In a possible design, the method in the third aspect can further include that the second device receives first information from the first device, where the first information indicates a method for determining the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task; and the second device determines the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task according to information of a third sample corresponding to the candidate participating entity as the first role, can include that the second device determines the candidate participating entity as the target participating entity of the first VFL task and / or the role of the candidate participating entity as the target participating entity in the first VFL task according to the information of the third sample corresponding to the candidate participating entity as the first role and the first information. In this way, the second device can also determine the target participating entity of the first VFL task and / or the role of the target participating entity based on the first information, i.e., the selection policy of the target participating entity and / or the corresponding role, issued by the first device, to more efficiently complete selection of the target participating entity and / or the role.

[0026] In a possible design, the first device can be a device that initiates the first VFL task. In embodiments of this application, the first device that sends the first information can also be a subscriber of a VFL service, can not have a VFL capability, or the first device can also be a device with a VFL capability, for example, a device that initiates the first VFL task or a primary participant or a secondary participant.

[0027] In a possible design, the second device receiving the first information from the first device can include that the second device receives a second request message from the first device, where the second request message includes the first information, and the second request message is used to request execution of the first VFL task. In this way, the first information can be carried in the second request message initiated by the first device to initiate the first VFL task, to reduce signaling overhead.

[0028] In a possible design, the method in the second aspect can further include: the first device sending first information to the second device.

[0029] In a possible design, the second request message can be used for the first device to request to participate in the first VFL task. In this way, in a case where the first device supports the first VFL task, the first device can request to participate in the first VFL task by multiplexing the second request message used by the first device to initiate the first VFL task, so as to reduce signaling overhead. It should be understood that the first device can indicate the request to participate in the first VFL task by using other defined or newly added messages. In the embodiments of the present application, the second request message is used to initiate the first VFL task, and can also be understood as being used to request the first VFL task or request the first VFL service, and the like, which are not limited herein.

[0030] In a possible design, the second request message can include information used to indicate that the first device requests to participate in the first VFL task. The information used to indicate that the first device requests to participate in the first VFL task can include at least one of the following: information about a role that the first device requests to assume in the first VFL task (or a role that the first device expects to assume in the first VFL task), information about a role that the first device supports to assume in the first VFL task, or information about a sample that the first device supports as a different role. It should be understood that in some embodiments, the information used to indicate that the first device requests to participate in the first VFL task can also be carried in other messages and sent, which is not limited herein.

[0031] In a possible design, the first information can include information used to indicate that the first device requests to participate in the first VFL task. In this way, the first device requesting to participate in the first VFL task can be sent as a selection target participating entity strategy in the form of the first information.

[0032] In a possible design, the information about the candidate participating entity that supports the first VFL task as the third sample corresponding to the first role can include information about the first device as the third sample corresponding to the first role. In this way, in a case where the first device is a candidate participating entity, the information about the candidate participating entity as the third sample corresponding to the first role acquired by the second device includes the information about the first device as the third sample corresponding to the first role, and at this time, the first role can be a role requested by the second device to participate in the first VFL task or a role supported by the second device in the first VFL task, so as to facilitate the second device to determine whether the first device can be a target participating entity in the future, and in a case where the first device is a target participating entity, the sample information can also be used to perform sample alignment, so that each target participating entity can perform model training and / or inference related to the first VFL task based on sample data after sample alignment.

[0033] In a possible design, the information for indicating that the first device requests to participate in the first VFL task can include information of a sample corresponding to the first device as the first role, so as to reduce signaling overhead.

[0034] In a possible design, the method in the second aspect can further include: sending, by the second device, information of a target participating entity participating in the first VFL task and / or information of a role of the target participating entity in the first VFL task to the first device.

[0035] In a possible design, the method in the second aspect can further include: receiving, by the first device, information of a target participating entity participating in the first VFL task and / or information of a role of the target participating entity in the first VFL task from the second device.

[0036] In a possible design, the second device can obtain the information of the third sample corresponding to the candidate participating entity as the first role, which can include: sending, by the second device, a first request message to the candidate participating entity, where the first request message is used to request to obtain the information of the sample corresponding to the candidate participating entity as the first role. Receiving, by the second device, a first response message from the candidate participating entity, where the first response message includes the information of the third sample corresponding to the candidate participating entity as the first role. In this way, the second device can obtain the information of the third sample corresponding to the candidate participating entity as the first role by sending the first request message to the candidate participating entity.

[0037] In a fourth aspect, a communication device is provided for implementing the methods described above. The communication device can be the first device in the first aspect, or a device containing the first device, or a device contained in the first device, such as a chip. The communication device includes corresponding modules, units, or means for implementing the methods described in the first aspect, which can be implemented by hardware, software, or by executing corresponding software by hardware. The hardware or software includes one or more modules or units corresponding to the functions described above.

[0038] In some possible designs, the communication device includes a processing module and a communication module. The communication module is configured to receive information of a first sample corresponding to a candidate participating entity supporting a first vertical federated learning (VFL) task as a first role, where the communication device is a device initiating the first VFL task. The processing module is configured to determine a target participating entity of the first VFL task and / or a role of the target participating entity in the first VFL task according to the information of the first sample corresponding to the candidate participating entity as the first role.

[0039] In a possible design, the communication module is further configured to send, to the second device, information about a target participant entity participating in the first VFL task and / or information about a role of the target participant entity in the first VFL task.

[0040] In a possible design, the role of the target participant entity in the first VFL task can be a primary participant or a secondary participant.

[0041] In a possible design, the first sample is an intersection of a second sample and a third sample corresponding to the candidate participant entity as the first role, and the second sample is a sample supported by the communication device.

[0042] In a possible design, the communication module is further configured to send, to the second device, information about a sample used by the target participant entity to perform the first VFL task.

[0043] In a possible design, the target participant entity of the first VFL task and / or the role of the target participant entity in the first VFL task can be determined according to at least one of the following: a weight of a feature of a sample corresponding to a role supported by the candidate participant entity, a membership weight corresponding to the candidate participant entity, a membership corresponding to the candidate participant entity, or a role supported by the candidate participant entity in the first VFL.

[0044] In a possible design, the communication module can include a receiving module and a sending module. The sending module is configured to implement the sending function of the communication device in the fourth aspect, and the receiving module is configured to implement the receiving function of the communication device in the fourth aspect.

[0045] In a possible design, the communication device in the fourth aspect can further include a storage module that stores programs or instructions. When the processing module executes the programs or instructions, the communication device in the fourth aspect can execute the method in the first aspect.

[0046] In a fifth aspect, a communication device is provided for implementing the methods described above. The communication device can be the second device in the second aspect, or a device including the second device, or a device included in the second device, such as a chip. The communication device includes corresponding modules, units, or means for implementing the methods in the second aspect, which can be implemented by hardware, software, or by hardware executing corresponding software. The hardware or software includes one or more modules or units corresponding to the functions described above.

[0047] In some possible design, the communication apparatus includes a processing module and a communication module. The communication module is configured to obtain information of a candidate participant entity supporting a first VFL task as a third sample corresponding to a first role. The processing module is configured to determine information of the candidate participant entity as a first sample corresponding to the first role according to information of a second sample supported by a first apparatus and the information of the candidate participant entity as the third sample corresponding to the first role, the first sample being an intersection of the second sample and the third sample, and the first apparatus being an apparatus initiating the first VFL task. The communication module is further configured to send the information of the candidate participant entity as the first sample corresponding to the first role to the first apparatus.

[0048] In a possible design, the communication module is further configured to receive information of a target participant entity participating in the first VFL task and / or information of a role of the target participant entity in the first VFL task from the first apparatus.

[0049] In a possible design, the information of the target participant entity participating in the first VFL task and / or the information of the role of the target participant entity in the first VFL task is determined according to the information of the candidate participant entity as the first sample corresponding to the first role.

[0050] In a possible design, the role of the target participant entity in the first VFL task is a primary participant or a secondary participant.

[0051] In a possible design, the communication module configured to obtain the information of the candidate participant entity supporting the first VFL task as the third sample corresponding to the first role can include: the communication module configured to send a first request message to the candidate participant entity, the first request message being used to request the candidate participant entity to obtain information of a sample corresponding to the first role. The communication module is further configured to receive a first response message from the candidate participant entity, the first response message including the information of the third sample corresponding to the first role.

[0052] In a possible design, the communication module can include a receiving module and a sending module. The sending module is configured to implement the sending function of the communication apparatus in the fifth aspect, and the receiving module is configured to implement the receiving function of the communication apparatus in the fifth aspect.

[0053] In a possible design, the communication apparatus in the fifth aspect can further include a storage module storing a program or an instruction. When the processing module executes the program or the instruction, the communication apparatus in the fifth aspect can execute the method in the second aspect.

[0054] In a sixth aspect, a communication apparatus is provided for implementing the various methods described above. The communication apparatus can be the second device in the third aspect above, or a device including the second device, or a device included in the second device, such as a chip. The communication apparatus includes corresponding modules, units, or means for implementing the methods of the third aspect above, which can be implemented by hardware, software, or by executing corresponding software by hardware. The hardware or software includes one or more modules or units corresponding to the functions described above.

[0055] In some possible designs, the communication apparatus includes a processing module and a communication module. The communication module is configured to obtain information of a candidate participating entity supporting a first VFL task as a third sample corresponding to a first role. The processing module is configured to determine a target participating entity of the first VFL task and / or a role of the target participating entity in the first VFL task according to the information of the candidate participating entity supporting the first VFL task as the third sample corresponding to the first role.

[0056] In one possible design, the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task can be determined according to at least one of the following: a weight of a feature of a sample corresponding to a role supported by the candidate participating entity, a membership weight corresponding to the candidate participating entity, a membership corresponding to the candidate participating entity, or a role supported by the candidate participating entity in the first VFL.

[0057] In one possible design, the communication module is further configured to receive first information from the first device, where the first information indicates a method of determining the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task. The processing module configured to determine the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task according to the information of the candidate participating entity supporting the first VFL task as the third sample corresponding to the first role can include that the processing module is configured to determine the candidate participating entity as the target participating entity of the first VFL task and / or the role of the candidate participating entity as the target participating entity in the first VFL task according to the information of the candidate participating entity supporting the first VFL task as the third sample corresponding to the first role and the first information.

[0058] In one possible design, the communication module configured to receive the first information from the first device can include that the communication module is configured to receive a second request message from the first device, where the second request message includes the first information, and the second request message is used to request to perform the first VFL task.

[0059] In one possible design, the second request message can be further used to request the first device to participate in the first VFL task.

[0060] In a possible design, the second request message can include information used to indicate that the first device requests to participate in the first VFL task.

[0061] In a possible design, the first information can include information used to indicate that the first device requests to participate in the first VFL task.

[0062] In a possible design, the information of the candidate participating entity supporting the first VFL task as the third sample corresponding to the first role can include information of the first device as the third sample corresponding to the first role.

[0063] In a possible design, the information used to indicate that the first device requests to participate in the first VFL task can include information of the first device as the third sample corresponding to the first role.

[0064] In a possible design, the communication module is further configured to send, to the first device, information of a target participating entity participating in the first VFL task and / or information of a role of the target participating entity in the first VFL task.

[0065] In a possible design, the communication module configured to obtain the information of the candidate participating entity supporting the first VFL task as the third sample corresponding to the first role can include: a communication module configured to send, to the candidate participating entity, a first request message used to request to obtain information of the candidate participating entity as a sample corresponding to the first role. The communication module is further configured to receive a first response message from the candidate participating entity, where the first response message includes the information of the candidate participating entity as the third sample corresponding to the first role.

[0066] In a possible design, the communication module can include a receiving module and a sending module. The sending module is configured to implement the sending function of the communication device in the sixth aspect, and the receiving module is configured to implement the receiving function of the communication device in the sixth aspect.

[0067] In a possible design, the communication device in the sixth aspect can further include a storage module that stores programs or instructions. When the processing module executes the programs or instructions, the communication device in the sixth aspect can execute the method in the third aspect.

[0068] In a seventh aspect, a communication device (for example, the communication device can be a chip or a chip system) is provided. The communication device includes a processor configured to implement the functions involved in any of the above aspects.

[0069] In a possible design, the communication apparatus further includes a memory configured to store necessary program instructions and data. The processor is coupled to the memory, and is configured to execute the computer program or instructions stored in the memory, so that the communication apparatus performs the method in any possible implementation of the first aspect to the third aspect.

[0070] In a possible design, the communication apparatus in the seventh aspect further includes a transceiver. The transceiver can be a transceiver circuit or an interface circuit. The transceiver can be configured to enable the communication apparatus in the seventh aspect to communicate with another communication apparatus.

[0071] In a possible design, the processor can be integrated with the memory.

[0072] In some possible designs, when the apparatus is a chip system, the apparatus can be formed by a chip, or can include a chip and other discrete devices.

[0073] The eighth aspect provides a communication apparatus, including a processor and an interface circuit. The interface circuit is configured to receive a signal from another communication apparatus outside the communication apparatus and transmit the signal to the processor, or send a signal from the processor to another communication apparatus outside the communication apparatus. The processor is configured to implement the method in any possible implementation of the first aspect to the third aspect by means of a logic circuit or by executing code instructions.

[0074] It can be understood that, when the communication apparatus in any one of the seventh aspect or the eighth aspect is a chip, the sending action / functionality described above can be understood as output, and the receiving action / functionality described above can be understood as input.

[0075] The ninth aspect provides a communication chip, in which instructions are stored. When the chip is run on a communication device, the instructions cause the method in any one of the first aspect to the third aspect to be implemented.

[0076] The tenth aspect provides a computer-readable storage medium, in which computer programs or instructions are stored. When the computer programs or instructions are run on a communication apparatus, the communication apparatus can perform the method in any one of the first aspect to the third aspect.

[0077] The eleventh aspect provides a computer program product including instructions. When the computer program code is run on a communication apparatus, the communication apparatus can perform the method in any one of the first aspect to the third aspect.

[0078] The twelfth aspect provides a communication system, including a first apparatus configured to implement the method in the first aspect, and a second apparatus configured to implement the method in the second aspect. BRIEF DESCRIPTION OF DRAWINGS

[0079] FIG. 1 is a schematic diagram of a network architecture suitable for use with embodiments of the present application;

[0080] FIG. 2 is a flow diagram of a method for determining VFL participating entities;

[0081] FIG. 3 is a flow diagram of another method for determining VFL participating entities;

[0082] FIG. 4 is a schematic diagram of an architecture of a communication system according to an embodiment of the present application;

[0083] FIG. 5 is a flow diagram of a communication method according to an embodiment of the present application;

[0084] FIG. 6 is a flow diagram of another communication method according to an embodiment of the present application;

[0085] FIG. 7 is a flow diagram of yet another communication method according to an embodiment of the present application;

[0086] FIG. 8 is a flow diagram of still another communication method according to an embodiment of the present application;

[0087] FIG. 9 is a schematic diagram of a communication apparatus according to an embodiment of the present application;

[0088] FIG. 10 is a schematic diagram of another communication apparatus according to an embodiment of the present application. DETAILED DESCRIPTION

[0089] Embodiments of the present application will be presented in relation to systems that can include a plurality of devices, components, modules, and the like. It should be understood and appreciated that each system can include additional devices, components, modules, and the like and / or can not include all of the devices, components, modules, and the like discussed in relation to the figures. Further, a combination of these approaches can be used.

[0090] The technical solutions of the embodiments of the present application can be applied to various communication systems, for example, a wireless fidelity (Wi-Fi) system, a vehicle to everything (V2X) communication system, a device to device (D2D) communication system, a vehicle-to-everything communication system, a worldwide interoperability for microwave access (WiMAX) communication system, a 4th generation (4G) mobile communication system, such as a long term evolution (LTE) system, a 5th generation (5G) mobile communication system, such as a new radio (NR) system, and a future communication system.

[0091] In a communication system, a part operated by an operator can be referred to as a public land mobile network (PLMN), and can also be referred to as an operator network, and the like. The PLMN is a network established and operated by a government or an operator approved thereby for the purpose of providing a public land mobile communication service, and is mainly a public network in which a mobile network operator (MNO) provides a mobile broadband access service for users. The PLMN described in the embodiments of the present application can be specifically a network conforming to the requirements of the 3GPP standard, referred to as a 3GPP network. The 3GPP network usually includes, but is not limited to, a 5G network, a 4G network, and other future communication systems.

[0092] Referring to FIG. 1, FIG. 1 is a schematic diagram of a network architecture suitable for the embodiments of the present application, taking a 5G network architecture based on a service-based architecture (SBA) in a non-roaming scenario defined in the 3GPP standardization process as an example. As shown in FIG. 1, the network architecture can include a terminal device, a radio (R) access network (AN), and a core network (CN). The terminal device accesses a data network (DN) through an access network device and a core network device.

[0093] The terminal device can be a terminal device with transceiver function, or a chip or chip system that can be disposed in the terminal device. The terminal device can also be referred to as a user equipment (UE), an access terminal, a subscriber unit, a user station, a mobile station (MS), a mobile station, a remote station, a remote terminal, a mobile device, a user terminal, a terminal, a wireless communication device, a user agent, or a user apparatus. The terminal device in the embodiments of the present application can be a mobile phone, a cellular phone, a smart phone, a Pad, a wireless data card, a personal digital assistant computer (PDA), a wireless modem, a handset, a laptop computer, a machine type communication (MTC) terminal, a computer with wireless transceiver function, a virtual reality (VR) terminal, an augmented reality (AR) terminal, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical treatment, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in smart city, a wireless terminal in smart home, a vehicle-mounted terminal, a road side unit (RSU) with terminal function, etc. The terminal device in the present application can also be a vehicle-mounted module, a vehicle-mounted module, a vehicle-mounted component, a vehicle-mounted chip, or a vehicle-mounted unit built into a vehicle as one or more components or units.

[0094] The (R)AN is used to implement access-related functions, can provide network access functions for authorized users in a specific area, and can determine transmission links of different qualities to transmit user data according to the level of a user, the demand of a service, and the like. The (R)AN forwards control signals and user data between a terminal device and a CN. The (R)AN can include one or more access network devices, which can also be referred to as an access network node, a radio access network (R)AN node, a (R)AN entity, or an access node, and the like, to help a terminal device implement wireless access and have a wireless transceiving function. An interface between an access network device and a terminal device can be a Uu interface (or an air interface, that is, messages exchanged between an access network device and a terminal device can be referred to as air interface messages). Of course, in future communications, the interface name can remain unchanged, or can be replaced by another name, which is not limited in this application.

[0095] The access network device includes, but is not limited to, a base station, an evolved Node B (eNodeB / eNB), an access point (AP), a transmission reception point (TRP or transmission point, TP), a next generation NodeB (gNB), a base station in a future mobile communication system, or an access node in a Wi-Fi system, etc. The access network device can be a macro base station, a micro base station or an indoor station, a relay node or a donor node, an open radio access network (ORAN), or a wireless controller in a centralized radio access network (CRAN) scenario. The access network device can also be one or a group (including multiple antenna panels) of antenna panels of a base station in 5G, or can also be a network node constituting a gNB, a TRP or a TP or a transmission measurement function (TMF), such as a central unit (CU), a distributed unit (DU), a CU-control plane (CP), a CU-user plane (UP), or a radio unit (RU), an RSU with base station function, and optionally, the access network device can also be a server, a wearable device, a vehicle or a vehicle-mounted device, etc. For example, the access network device in V2X technology can be an RSU. All or part of the functions of the network device in this application can also be implemented by software functions running on hardware, or by virtualized functions instantiated on a platform (such as a cloud platform). The access network device in this application can also be a logical node, a logical module or software that can implement all or part of the functions of the access network device.

[0096] The CU and the DU can be separately arranged, or can also be included in the same network element, for example, a baseband unit (BBU). The RU can be included in a radio frequency device or a radio frequency unit, for example, included in a remote radio unit (RRU), an active antenna unit (AAU), or a remote radio head (RRH). It can be understood that the access network device can be a CU node, or a DU node, or a device including the CU node and the DU node. In addition, the CU can be divided into a network device in the access network RAN, or the CU can be divided into a network device in the CN, which is not limited herein.

[0097] The CU (or CU-CP and CU-UP), DU, or RU can also have different names in different systems, but those skilled in the art can understand their meanings. For example, in an ORAN system, the CU can also be referred to as an O-CU (open CU), the DU can also be referred to as an O-DU, the CU-CP can also be referred to as an O-CU-CP, the CU-UP can also be referred to as an O-CU-UP, and the RU can also be referred to as an O-RU. For the convenience of description, the CU, the CU-CP, the CU-UP, the DU, and the RU are taken as examples for description in this application. Any one of the CU (or CU-CP, CU-UP), the DU, and the RU in this application can be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0098] The access network device and the terminal device can be fixed in position or movable. The access network device and the terminal device can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; can also be deployed on the water surface; and can also be deployed on an airplane, a balloon, and a man-made satellite in the air. Embodiments of this application do not limit the application scenarios of the access network device and the terminal device.

[0099] The CN can include, but is not limited to, the following network functions (NFs): a user plane function (UPF), an authentication server function (AUSF), an access and mobility management function (AMF), a session management function (SMF), a network slice selection function (NSSF), a network exposure function (NEF), a network repository function (NRF), a policy control function (PCF), a unified data management (UDM), a unified data repository (UDR), a network data analytics function (NWDAF), and an application function (AF).

[0100] The NF functions included in the CN are further briefly described below.

[0101] 1. UPF: mainly responsible for user data processing (forwarding, receiving, charging, etc.). For example, the UPF can receive user data from the DN, and forward the user data to the terminal through the access network device. The UPF can also receive user data from the terminal device through the access network device, and forward the user data to the DN. In a protocol data unit (PDU) session, the UPF directly connected with the DN through N6 is also called a protocol data unit session anchor (PSA).

[0102] 2. AUSF: mainly used for performing security authentication of the terminal device.

[0103] 3. AMF: mainly used for mobility management in the mobile network. For example, user location update, user registration network, user handover, etc.

[0104] 4. SMF: mainly used for session management in mobile networks. For example, session establishment, modification, release. Specific functions include, for example, assigning an internet protocol (IP) address to a user, selecting a UPF that provides packet forwarding functions, etc.

[0105] 5. NSSF: mainly used for selecting network slices for terminal devices.

[0106] 6. NEF: mainly used for supporting the opening of capabilities and events. For example, the NEF can expose some capabilities of the 5G network to third-party applications through an application program interface (API), and the third-party application can obtain some capabilities of the 5G network by calling the API provided by the NEF through the AF, so that the third-party application can control some behaviors of the 5G network and terminal devices.

[0107] 7. NRF: mainly used for providing network element discovery functions, providing network element information corresponding to network element types based on requests from other network elements, and also providing network element management services such as network element registration, update, deregistration, and network element state subscription and push.

[0108] 8. PCF: mainly used to support providing a unified policy framework to control network behavior, providing policy rules to control layer network functions, and also responsible for obtaining user subscription information related to policy decision. The PCF can provide policies such as quality of service (QoS) policies, slice selection policies, etc. to the AMF and SMF.

[0109] 9. UDM: mainly used for storing user data, such as subscription data, authentication / authorization data, etc.

[0110] 10. UDR: mainly used for storing structured data, including subscription data and policy data, externally exposed structured data, and application-related data.

[0111] 11、NWDAF: mainly used for collecting data (including one or more of terminal device data, access network device data, core network element data, and third-party application device data), wherein the data can be data of the terminal device, the access network device, the core network element, or the third-party application device itself, or data of the terminal device on the access network device, the core network element, or the third-party application device, then performing data analysis according to the collected data, and outputting data analysis results for network, network management device, and application to execute policy decision. The NWDAF can use a machine learning model to perform data analysis. 3GPP splits the training function and the inference function of the NWDAF, and one NWDAF can support only the model training function, or only the data inference function, or both the model training function and the data inference function. Among them, the NWDAF supporting the model training function can also be called a training NWDAF, or a NWDAF supporting a model training logical function (MTLF) (referred to as MTLF). The training NWDAF can perform model training according to the obtained data to obtain a trained model. The NWDAF supporting the data inference function can also be called an inference NWDAF, or a NWDAF supporting an analytics logical function (AnLF) (referred to as AnLF). The inference NWDAF can input input data into the trained model to obtain analysis results or inference data. In the embodiments of the present application, the training NWDAF refers to the NWDAF supporting at least the model training function. As a possible implementation method, the training NWDAF can also support the data inference function. The inference NWDAF refers to the NWDAF supporting at least the data inference function. As a possible implementation method, the inference NWDAF can also support the model training function. If a NWDAF supports both the model training function and the data inference function, the NWDAF can be called a training NWDAF, an inference NWDAF, or a training and inference NWDAF, or a NWDAF. In the embodiments of the present application, one NWDAF can be a separate network element, or can be combined with other network elements, for example, the NWDAF is set in the PCF or the AMF.

[0112] 12、AF: transmits application-side requirements for the network side, such as QoS requirements or user state event subscription, etc. The AF can be a third-party functional entity, or an application server deployed by an operator.

[0113] It should be understood that the above exemplary shows several core network elements contained in the CN, in addition to which, other core network elements can also be included, for example, in a machine learning (ML) scenario, the above CN can also include an analytics data repository function (ADRF) (not shown in FIG. 1), which can be used to store model-related data. Among them, the data can be generated by AnLF, and the ADRF can provide the model-related data to the MTLF according to the request of the MTLF.

[0114] It can be understood that the above-mentioned network elements or functions can be both a physical entity in a hardware device and a software instance running on a dedicated hardware, or a virtualized function instantiated on a shared platform (for example, a cloud platform). In short, an NF can be implemented by hardware or software.

[0115] The DN is a network located outside the operator network, and the operator network can access multiple DNs. Various services can be deployed on the DN, and data and / or voice services can be provided for terminal devices. For example, the DN is a private network of a certain smart factory, and the sensors installed in the workshop of the smart factory can be terminal devices. A control server of the sensors is deployed in the DN, and the control server can provide services for the sensors. The sensors can communicate with the control server, obtain instructions from the control server, and transmit the collected sensor data to the control server according to the instructions. For another example, the DN is an internal office network of a certain company, and the mobile phones or computers of employees of the company can be terminal devices. The mobile phones or computers of the employees can access information and data resources on the internal office network of the company.

[0116] In FIG. 1, Nnef, Nnrf, Npcf, Nudm, Nudr, Nnwdaf, Naf, Nausf, Namf, Nsmf, N1, N2, N3, N4, and N6 are interface sequence numbers. Exemplarily, the meanings of the above interface sequence numbers can refer to the meanings defined in the 3GPP standard protocol, and the meanings of the above interface sequence numbers are not limited in the present application. It should be noted that the interface names between the network functions in FIG. 1 are only an example, and the interface names of the system architecture can also be other names in the specific implementation, which are not limited in the present application. In addition, the names of the messages (or signaling) transmitted between the above network elements are also only an example, and do not constitute any limitation on the functions of the messages themselves.

[0117] It should be noted that in the architecture shown in FIG. 1, the interface between the (R)AN and the CN can also be referred to as an NG interface (not shown in FIG. 1), and the (R)AN and the CN are connected through the NG interface. The NG interface can include an NG-C interface and an NG-U interface, where the NG-C interface is a control plane interface, the connection parties are the (R)AN and the AMF, and is used to transmit control plane data; and the NG-U interface is a user plane interface, the connection parties are the (R)AN and the UPF, and is used to transmit user plane data.

[0118] It should be understood that the AMF, SMF, UPF, NEF, AUSF, NRF, PCF, and UDM shown in FIG. 1 can be understood as network elements in the core network for implementing different functions, for example, can be combined into a network slice as needed. These core network network elements can be independent devices respectively, or can be integrated into the same device to implement different functions, and the specific form of the network elements is not limited in the present application.

[0119] It should also be understood that the above naming is only defined for the purpose of distinguishing different functions, and should not constitute any limitation on the present application. The present application does not exclude the possibility of using other names in 5G networks and future other networks. For example, in future networks, part or all of the above network elements can use the terms in 5G, or other names, etc.

[0120] For ease of understanding, the following will first introduce the related terms, concepts or technologies that may be involved in the embodiments of the present application:

[0121] 1. Federated Learning

[0122] Federated learning (FL) is a distributed machine learning method in which multiple participants interact model parameters through a secure mechanism without interacting or sharing original training data, so as to achieve a collaborative training effect. In other words, federated learning is an encrypted distributed machine learning technology. Federated learning can fully utilize the data and computing power of participants, so that multiple parties can collaboratively build a general and robust machine learning model without sharing data. Therefore, federated learning can effectively help multiple institutions to meet the requirements of user privacy protection, data security and government regulations, and to use data and build models. In a word, federated learning is to share knowledge and parameters without exchanging any own data.

[0123] Federated learning includes horizontal federated learning (HFL) and VFL. In the context of increasingly strict data regulation, federated learning can solve key problems such as data ownership, data privacy, data access, and access to heterogeneous data. A row of a data matrix horizontally represents a training sample, and a column of the data matrix vertically represents a data feature. HFL is federated learning of multiple participants with multiple rows of samples with the same features, that is, the training data of each participant is horizontally divided. HFL is also called feature-aligned federated learning, that is, the data features of the participants are aligned, and HFL can increase the total amount of training samples. VFL, as a machine learning technology, can be used to solve model training and inference in the case where each member is unwilling to share original data, and is suitable for the case where the training sample IDs of participants overlap more, and the data features of the participants overlap less. VFL is federated learning of different data features of common samples of multiple participants, that is, the training data of each participant is vertically divided, so it is called vertical federated learning. VFL is federated learning of different data features of common samples of multiple participants, that is, the training data of each participant is vertically divided. VFL is also called sample-aligned federated learning, that is, the training samples of the participants are aligned, and VFL can increase the dimension of the training data features.

[0124] 2、analysis ID

[0125] The analysis ID can be used to indicate an analysis service or an analysis service, simply referred to as a service. The service is related to a model, that is, the model can be used to perform the service. Or described as, the analysis ID is related to a model, that is, the model is used to perform the service corresponding to the analysis ID.

[0126] Or it can also be understood that the MTLF is related to the analysis ID, that is, the model provided by the MTLF supports the execution of the service corresponding to the analysis ID. Illustratively, the MTLF can be related to one or more analysis IDs. It can be understood that the MTLF can provide a model for each analysis ID in the one or more analysis IDs. For example, MTLF1 is related to analysis ID1 and analysis ID2, that is, MTLF1 corresponds to analysis ID1 and analysis ID2, then MTLF1 can provide a model for the service corresponding to analysis ID1, and a model for the service corresponding to analysis ID2.

[0127] 3、interoperability indicator

[0128] Exemplarily, the interoperability indicator can correspond to the MTLF, or correspond to the analytics indicator, or correspond to the analytics indicator of the MTLF. Alternatively, the interoperability indicator is related to the MTLF, or the interoperability indicator is related to the analytics indicator. Wherein, the interoperability indicator can also be referred to as interoperability indication, or ML model interoperability indicator, or model interoperability indicator (ML Model interoperability indicator).

[0129] The interoperability indicator includes a list of vendors, or alternatively described as a list of NWDAF providers (or suppliers). The vendors in the list of vendors are allowed to retrieve or use the models provided by the MTLF. The interoperability indicator also indicates that the MTLF supports the NWDAF request of the models provided by the MTLF for the vendors in the list of vendors. The interoperability indicator also indicates that the vendors in the list of vendors are allowed to obtain the models from the MTLF. The interoperability indicator also indicates that the MTLF allows the vendors in the list of vendors to obtain the models from the MTLF.

[0130] The interoperability identification provides a list of MTLF providers, for example, the interoperability identification represents a vendor identification, or the interoperability identification is associated with a vendor identification. The interoperability identification can be associated with an analytics identification, for example, one-to-one, indicating that the MTLF corresponding to the interoperability identification allows the corresponding vendor or the MTLF included in the vendor to obtain the model corresponding to the analytics identification, and / or, indicating that the MTLF corresponding to the interoperability identification is allowed to interoperate the model corresponding to the analytics identification with the AnLF. Optionally, one MTLF can have one or more interoperability identifications, and if there are multiple interoperability identifications, the multiple interoperability identifications correspond to different analytics identifications respectively. For example, MTLF NF ID 1 corresponds to analytics identification 1 and analytics identification 2, wherein the MTLF to which the MTLF NF ID 1 belongs has interoperability identification 1 and interoperability identification 2, interoperability identification 1 corresponds to analytics identification 1, and interoperability identification 2 corresponds to analytics identification 2. Optionally, if the MTLF to which the MTLF NF ID 1 belongs and the MTLF to which the MTLF NF ID 2 belongs belong to support interoperability, the MTLF to which the MTLF NF ID 2 belongs can have interoperability identification 1 and interoperability identification 2, wherein interoperability identification 1 corresponds to analytics identification 1, and / or interoperability identification 2 corresponds to analytics identification 2, that is, the MTLF of the same vendor can have the same interoperability identification for the same analytics identification. In addition, if the MTLF to which the MTLF NF ID 1 belongs and the MTLF to which the MTLF NF ID 2 belongs belong to different vendors, the MTLF to which the MTLF NF ID 2 belongs can have interoperability identification 3 and interoperability identification 4, wherein interoperability identification 3 corresponds to analytics identification 1, and / or interoperability identification 4 corresponds to analytics identification 2, that is, the MTLF of the same vendor can have different interoperability identifications for the same analytics identification.

[0131] For example, analytics identification 1 is associated with model 1, that is, model 1 is used to perform the service corresponding to analytics identification 1, and analytics identification 1 is associated with interoperability identification 1, that is, model 1 is associated with interoperability identification 1. It is assumed that interoperability identification 1 includes the identification of vendor 1 and the identification of vendor 2, that is, model 1 can be provided for use by vendor 1 and vendor 2, or it can be understood that if the vendor of the NWDAF is vendor 1 or vendor 2, the NWDAF can use model 1.

[0132] For example, one MTLF can have one or more interoperability identifications, and if there are multiple interoperability identifications, the multiple interoperability identifications can correspond to different analytics identifications. For example, MTLF 1 corresponds to analytics identification 1 and analytics identification 2, wherein interoperability identification 1 corresponds to analytics identification 1, and interoperability identification 2 corresponds to analytics identification 2.

[0133] 4、Model producer

[0134] The model producer is a subject that produces a model, or a subject that has the right to provide model information to other entities according to network configuration, and the consumer can obtain the model according to the model information. The model information includes but is not limited to: a uniform resource locator (URL) of a model file, a model itself, and the like.

[0135] 5. Vendor ID

[0136] The vendor ID of the vendor, which can also be represented as network element information, identifies the vendor or manufacturer of the network element. For example, Vendor ID1 identifies the vendor of the NWDAF network element.

[0137] The vendor ID can be used to identify a device manufacturer. One vendor ID can correspond to one or more NWDAF device IDs. For example, NWDAF NF ID 1 and NWDAF NF ID 2 can both correspond to vendor ID 1, i.e., the MTLF to which MTLF NF ID 1 belongs and the MTLF to which MTLF NF ID 2 belongs belong to the same vendor, and the vendor ID of the vendor is vendor ID 1.

[0138] The above briefly describes the terms involved in the present application, which will not be repeated in the following embodiments. In addition, the above description of the terms is only for the convenience of understanding and does not limit the protection scope of the embodiments of the present application.

[0139] Currently, the 3rd generation partnership project (3rd generation partnership project, 3GPP) supports training and / or inference through vertical federated learning tasks to implement mobile network-related analysis services. Before model training, the participating entities participating in VFL model training and / or the roles of the participating entities in VFL model training need to be determined.

[0140] For example, FIG. 2 is a flowchart of a method for determining VFL participating entities, in which the NWDAF is a VFL coordinator NF and an active participant, and the AF is a passive VFL participant. As shown in FIG. 2, the method includes the following steps, and the parts not described in detail can refer to the existing protocol. Steps S201-S203 are the registration process of the AF, steps S204-S206 are the discovery process of the AF, and steps S207-S210 are the selection process of the AF determined by the NWDAF to perform VFL.

[0141] S201, the AF sends a registration request message to the NRF. Correspondingly, the NRF receives the registration request message from the AF.

[0142] The registration request message is used to request registration to the network, and the registration request includes the NF profile of the AF, which includes the VFL profile of the AF, used to indicate the VFL capability of the AF, and the VFL profile of the AF includes one or more of the following: AF NF type, Application Name, analytics ID(s), address information of the AF, service area, FL capability type information (such as FL server), or time interval of AF supporting FL. For example, the registration request message can be a Nnrf_NFManagement_NFRegister_request message.

[0143] It should be understood that if the AF is not in the operator's domain, the AF can be registered on the NRF through the NEF service (such as authentication and authorization), so as to send its NF profile including the VFL profile to the NRF through the NEF.

[0144] S202, the NRF stores the VFL profile.

[0145] S203, the NRF sends a registration response message to the AF. Correspondingly, the AF receives the registration response message from the NRF.

[0146] The registration response message is a response message of the registration request message.

[0147] S204, the NWDAF sends a discovery request message to the NRF. Correspondingly, the NRF receives the discovery request message from the NWDAF.

[0148] The discovery request message is used to request the NRF to discover the AF participating in the VFL. The discovery request message can include VFL participant information, which can include one or more of the following: indication of the capability of the AF required to support as a VFL client, ML model analytics ID(s), ML model filter such as Application ID(s). Optionally, the VFL participant information can also include a list of ML features used by the AF to train a local model.

[0149] For example, the discovery request message can be a Nnrf_NFDiscovery_Request message.

[0150] S205, the NRF authorizes.

[0151] The NRF authorizes the discovery request message Nnrf_NFDiscovery_Request in step S204. For example, the NRF discovers candidate AFs that can participate in the required VFL process according to the VFL participant information.

[0152] S206. The NRF sends a discovery response message to the NWDAF. Correspondingly, the NWDAF receives the discovery response message from the NRF.

[0153] Exemplarily, if the NRF authorizes, the NRF determines a matched NF instance set, i.e. candidate AF instances, according to the Nnrf_NFDiscovery_Request and the internal policy of the NRF, and delivers the NF profile of the candidate AF instances. Exemplarily, the discovery response message can be a Nnrf_NFDiscovery_Response message, and the NF profile of each NF instance can be sent to the NWDAF through the Nnrf_NFDiscovery_Response message.

[0154] S207. The NWDAF sends a VFL preparation request message to the candidate AF. Correspondingly, the candidate AF receives the VFL preparation request message from the NWDAF.

[0155] Exemplarily, the VFL preparation request message can be a VFL preparation request message, e.g. the NWDAF sends a VFL preparation request message to the candidate AF using the Naf_MLModelTraining_Subscribe or Naf_MLModelTrainingInfo_Request service with a ML preparation flag, checks whether each candidate AF can join the VFL training process, such as whether the candidate AF can meet the ML model training requirements (e.g. analysis ID, ML model interoperability information), data availability requirements (event ID list of local data for training, the data availability requirements can also include dataset statistical attributes, time window of data samples and minimum number of data samples), or availability time requirements (time span required by the VFL process), etc. The VFL preparation request message can include at least one of the following: analysis ID, role of VFL participant to perform VFL, VFL interoperability information, VFL joint ML training ID.

[0156] S208. The candidate AF determines whether to join the VFL.

[0157] Exemplarily, the candidate AF(s) checks whether it can meet the ML model training requirements, and / or, if the model information is provided in the VFL preparation request message in step S207, the candidate AF(s) also needs to check whether it can successfully download the model, and decides whether to join the VFL process according to the implementation. The example criteria used by the candidate AF(s) can be based on its availability, computing and communication capabilities, and ML model interoperability information.

[0158] S209. The candidate AF sends a VFL preparation response message to the NWDAF. Correspondingly, the NWDAF receives the VFL preparation response message from the candidate AF.

[0159] Exemplarily, the VFL preparation response message can be a VFL preparation response message, for example, the candidate AF(s) invokes a Naf_MLModelTraining_Subscribe response service operation or a Naf_MLModelTrainingInfo_Request response service operation to indicate whether to join the VFL, and if the VFL cannot be joined, the candidate AF(s) can carry a reason value in the VFL preparation response message, for example, the candidate AF(s) currently cannot support federated learning, or the candidate AF(s) is currently overloaded, etc.

[0160] S210. The NWDAF selects the AF participating in the VFL.

[0161] Exemplarily, the NWDAF selects or finally determines one or more AFs supporting the execution of federated learning according to the VFL preparation response message sent by one or more candidate AFs received in step S209.

[0162] Exemplarily, FIG. 3 is a flowchart of another method for determining VFL participating entities, wherein the AF is a VFL coordinator NF and an active participant, and the NWDAF is a passive VFL participant. As shown in FIG. 3, the method includes the following steps, and the parts not described in detail can refer to the existing protocol. Among them, steps S301-S303 are the registration process of the NWDAF, steps S304-S306 are the discovery process of the NWDAF, and steps S307-S310 are the selection process of the NWDAF for the AF to determine the execution of the VFL.

[0163] S301. The NWDAF sends a registration request message to the NRF. Correspondingly, the NRF receives the registration request message from the NWDAF.

[0164] The registration request message includes the NF profile of the NWDAF, and the NF profile includes a VFL profile for indicating the VFL capability thereof. The specific description of S301 can refer to the related description in S201 described above, and no further description is made.

[0165] S302. The NRF stores the VFL profile.

[0166] S303. The NRF sends a registration response message to the NWDAF. Correspondingly, the NWDAF receives the registration response message from the NRF.

[0167] S304, the AF sends a discovery request message to the NRF through the NEF. Correspondingly, the NRF receives the discovery request message from the AF through the NEF.

[0168] The discovery request message is used to request the NRF to discover the NWDAF participating in the VFL. The discovery request message can include VFL participant information, which can include one or more of the following: an indication of the capability of the NWDAF required to support as a VFL client, an analysis ID(s) of the ML model, and an ML model filter such as an Application ID(s). Optionally, the VFL participant information can also include a list of ML features used by the NWDAF to train a local model.

[0169] The discovery request message can be an example of an Nnrf_NFDiscovery_Request message.

[0170] S305, the NRF authorizes.

[0171] The discovery request message in step S304 of the NRF authorization is an Nnrf_NFDiscovery_Request. For example, the NRF discovers candidate NWDAFs that can participate in the required VFL process according to the VFL participant information.

[0172] S306, the NRF sends a discovery response message to the AF through the NEF. Correspondingly, the AF receives the discovery response message from the NRF through the NEF.

[0173] If the NRF authorizes, the NRF determines a matching NF instance set, i.e., candidate NWDAF instances, according to the Nnrf_NFDiscovery_Request and the internal policy of the NRF, and issues the NF profile of the candidate NWDAF instances. The discovery response message can be an example of an Nnrf_NFDiscovery_Response message, and the NF profile of each NF instance can be sent to the AF through the Nnrf_NFDiscovery_Response message.

[0174] S307, the AF sends a VFL preparation request message to the candidate NWDAF through the NEF. Correspondingly, the candidate NWDAF receives the VFL preparation request message from the NWDAF through the NEF.

[0175] Exemplarily, the VFL preparation request message can be a VFL preparation request message, e.g., the AF sends the VFL preparation request message to candidate AFs using Nnwdaf_MLModelTraining_Subscribe or Nnwdaf_MLModelTrainingInfo_Request service with ML preparation flag, checks whether each candidate NWDAF can join the VFL training process, e.g., whether the candidate NWDAF can meet the ML model training requirements (e.g., analytics ID, ML model interoperability information), available data requirements (event ID list of local data for training, the available data requirements can also include dataset statistical attributes, time window of data samples and minimum number of data samples), or availability time requirements (time span required by the VFL procedure), etc. Wherein, the VFL preparation request message can include at least one of the following: analytics ID, an indication of the role of the VFL participant to perform VFL, VFL interoperability information, VFL joint ML training ID.

[0176] S308, the candidate NWDAF determines whether to join the VFL.

[0177] Exemplarily, the candidate NWDAF(s) checks whether it can meet the ML model training requirements, and / or, if the model information is provided in the VFL preparation request message in step S307, the candidate NWDAF(s) also needs to check whether it can successfully download the model, and decides whether to join the VFL procedure according to the implementation. The example criteria used by the candidate NWDAF(s) can be based on its availability, computing and communication capabilities, and ML model interoperability information.

[0178] S309, the candidate NWDAF sends a VFL preparation response message to the AF. Correspondingly, the AF receives the VFL preparation response message from the candidate NWDAF.

[0179] S310, the AF selects the NWDAF participating in the VFL.

[0180] The specific implementation process of S309-S310 can refer to the related description in S209-S210 described above, and will not be described here.

[0181] As can be seen from the above, the NWDAF or AF as a coordinator only determines the final participants of the VFL based on the response message of the candidate participating entity feedback whether to join the VFL. However, different candidate participating entities support different sample information, and the sample information includes sample corresponding label and / or training data, and the candidate participating entity does not want to disclose its sample information.

[0182] For example, in the scenario shown in FIG. 2, when the analysis target is a UE, different AFs can have different sample data of the UE, one UE corresponds to one sample, and the sample identifier can be a UE ID or an identifier assigned by the AF corresponding to / associated with the UE ID, where the UE ID can be a subscription permanent identifier (SUPI) or a generic public subscription identifier (GPSI).

[0183] For different roles supported by the AF in the VFL, the sample data (training data and / or labels) provided by the AF for a certain UE sample is different, as shown in Table 1 below, A1-A4 are UE sample identifiers supported by AF1, B1-B4 are UE sample identifiers supported by AF2, and C1-C4 are UE sample identifiers supported by AF3. A1, B1, and C1 correspond to the same UE sample (corresponding to the same UE-SUPI, such as SUPI1), A2, B2, and C2 correspond to the same UE sample (corresponding to the same UE-SUPI, such as SUPI2), A3, B3, and C3 correspond to the same UE sample (corresponding to the same UE-SUPI, such as SUPI3), and A4, B4, and C4 correspond to the same UE sample (corresponding to the same UE-SUPI, such as SUPI3). However, the labels and / or training data corresponding to the same UE sample are different for different AFs, and the sample identifiers stored by AF1, AF2, and AF3 are different.

[0184] Table 1

[0185] As can be seen from Table 1 above, for AF1, it has sample data corresponding to A1-A3, but does not have sample data corresponding to A4. When AF1 supports as a slave participant in the VFL, it has training data corresponding to A1-A3, and when AF1 supports as a master participant, it does not have labels corresponding to A1 but has labels corresponding to A2-A3. For AF2, it has sample data corresponding to B1-B4, and when AF supports as a slave participant in the VFL, it has training data corresponding to B1-B4, and when AF supports as a master participant, it has labels corresponding to B1-B4. For AF3, it has sample data corresponding to C1-C3, but does not have sample data corresponding to C4. When AF3 supports as a slave participant in the VFL, it does not have training data corresponding to C1 but has training data corresponding to C2-C3, and when AF3 supports as a master participant, it has labels corresponding to C1-C3.

[0186] Therefore, in the above participant entity determination process, since the role information of the participant entity is introduced, it is impossible to determine the participant entity participating in the VFL and the role of the participant entity in the VFL task based on only the ACK of the feedback of the candidate participant entity whether the candidate participant entity can join the VFL, and it is also unclear how the role information affects the process of determining the participant entity participating in the VFL task and the role of the participant entity in the VFL task. To this end, the communication method provided in the embodiments of the present application can determine the participant entity participating in the VFL task and / or the role of the participant entity in the VFL task in the granularity of the role.

[0187] In order to better understand the embodiments of the present application, the following points are explained before the embodiments of the present application are introduced.

[0188] First, in the embodiments of the present application, "for indicating" can include direct indication and indirect indication. When describing that certain "indication information" is used to indicate A, it can include that the indication information directly indicates A or indirectly indicates A, and it does not mean that A must be carried in the indication information.

[0189] The information indicated by the indication information is referred to as to-be-indicated information. In the specific implementation process, there are many ways to indicate the to-be-indicated information, for example but not limited to, the to-be-indicated information can be directly indicated, such as the to-be-indicated information itself or an index of the to-be-indicated information. The to-be-indicated information can also be indirectly indicated by indicating other information, where the other information and the to-be-indicated information have an association relationship. The to-be-indicated information can also be indicated only by a part of the to-be-indicated information, and the other part of the to-be-indicated information is known or agreed in advance. For example, the indication of a specific information can also be achieved by means of the arrangement order of each information agreed in advance (for example, specified by a protocol), thereby reducing the indication overhead to a certain extent. At the same time, the common part of each information can be identified and indicated uniformly, so as to reduce the indication overhead caused by separately indicating the same information.

[0190] In addition, the specific indication manner can also be various existing indication manners, for example but not limited to, the above indication manners and various combinations thereof. The specific details of various indication manners can refer to the prior art, which will not be described herein. As can be seen from the above, for example, when multiple information of the same type needs to be indicated, the indication manners of different information can be different. In the specific implementation process, the required indication manner can be selected according to the specific needs, and the selected indication manner is not limited in the embodiments of the present application. In this way, the indication manner involved in the embodiments of the present application should be understood as covering various methods that can enable the to-be-indicated party to know the to-be-indicated information.

[0191] The to-be-indicated information can be sent as a whole or can be divided into multiple sub-information and sent separately, and the sending period and / or sending occasion of the sub-information can be the same or different. The specific sending method is not limited in the present application. The sending period and / or sending occasion of the sub-information can be predefined, for example, predefined according to a protocol, or configured by the transmitting end device by sending configuration information to the receiving end device.

[0192] Secondly, in the embodiments of the present application, the first, second and various numerical numbers are only for differentiation for convenience of description, and do not limit the scope of the embodiments of the present application. For example, different indication information is differentiated. For another example, the first indication information and the second indication information are only for differentiation of different indication information, and the sequence thereof is not limited. Those skilled in the art can understand that the words "first", "second" and the like do not limit the number and execution sequence, and the words "first", "second" and the like do not necessarily mean different.

[0193] Thirdly, in the embodiments of the present application, the descriptions such as "when", "in the case of", "if" and "whether" all mean that the device (such as the first device or the second device) will make corresponding processing under certain objective circumstances, which is not limited in time, and the device (such as the first device or the second device) does not necessarily have a judgment action when implemented, and it does not mean that there are other limitations.

[0194] Meanwhile, in the embodiments of the present application, the words "exemplary" or "for example" are used to mean by way of example, illustration or description. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of the present application should not be construed as more preferred or more advantageous than other embodiments or design schemes. Rather, the words "exemplary" or "for example" are used to present the relevant concept in a specific manner, for ease of understanding.

[0195] Finally, the network architecture and service scenarios described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that, with the evolution of network architecture and the appearance of new service scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0196] Please refer to FIG. 4, which is a schematic diagram of an architecture of a communication system to which the embodiments of the present application are applied. As an example, as shown in FIG. 4, the communication system includes a first device and a second device in communication with the first device.

[0197] The first device can have or not have a vertical federated learning capability, for example, the first device can request the second device to provide a vertical federated learning service or subscribe to a vertical federated learning service. For example, the second device can act as an initiator of a VFL task, and be configured to initiate or request to initiate a VFL task or a VFL service, or trigger a determination process of a participating entity of the VFL. The first device can be an NF type entity, device or network element, or can be an AF type entity, device or network element.

[0198] Optionally, the first device can also be a master participant or a slave participant of a vertical federated task. It can be understood that the first device is one of the members of the VFL task, and its role is a master participant or a slave participant. In one implementation, the first device is an initiator and a master participant of the VFL task.

[0199] The second device acts as a producer of a vertical federated service, for example, the second device can provide one or more of the following services: a vertical federated learning service, a vertical federated learning and inference service, a coordinator service, training participation, or inference participation, etc. For example, the second device can act as a coordinator of a VFL task, and be configured to determine or coordinate to determine a participating entity of the VFL task, and / or coordinate the participating entity to complete the VFL task, etc. The second device can be an NF type entity, device or network element, or can be an AF type entity, device or network element.

[0200] Optionally, the second device can also be a master participant or a slave participant of a vertical federated task. It can be understood that the second device is one of the members of the VFL task, and its role is a master participant or a slave participant. In one implementation, the second device is a coordinator and a master participant of the VFL task.

[0201] In the embodiments of the present application, the candidate participating entity or the participating entity can be a device, a device, a network element, or a node, etc. For example, it can be an NF type entity such as NWDAF, or an AF type entity, and can also be an access network device or a terminal device, which is not limited.

[0202] In one possible design, the first device can act as one of the participating entities of the VFL. In the embodiments of the present application, the first device can determine a target participating entity of a first vertical federated learning task and / or a role of the target participating entity in the first vertical federated learning task according to the information fed back by the second device that a candidate participating entity supporting the first vertical federated learning task is used as a first sample corresponding to the first role.

[0203] In a possible design, the second device can acquire information about candidate participating entities of the first longitudinal federated learning task as third samples corresponding to the first role, and determine a target participating entity participating in the first longitudinal federated learning task and / or a role of the target participating entity in the first longitudinal federated learning task according to the information about the candidate participating entities of the first longitudinal federated learning task as the third samples corresponding to the first role.

[0204] It should be understood that the communication system shown in FIG. 4 can also include other devices, apparatuses, network elements, and the like, such as an NRF, for discovering candidate participating entities, without limitation.

[0205] The communication method provided by the embodiments of the present application will be described in detail below with reference to FIGS. 5-8.

[0206] For example, FIG. 5 is a flowchart of a communication method provided by an embodiment of the present application, which is described by taking the communication between the first device and the second device shown in FIG. 4 as an example. Of course, the subject performing the actions of the first device in the method can also be a device / module in the first device, such as a chip, processor, processing unit, and the like in the first device, without limitation; and the subject performing the actions of the second device in the method can also be a device / module in the second device, such as a chip, processor, processing unit, and the like in the second device, without limitation.

[0207] As shown in FIG. 5, the communication method includes the following steps.

[0208] S501, the second device acquires information about candidate participating entities supporting a first VFL task as third samples corresponding to a first role.

[0209] The second device can have one or more of the following capabilities: the capability of acquiring information about samples of candidate participating entities, the capability of coordinating execution of the first VFL task, which can be understood as a coordinator in the first VFL task. The second device can be a network element or apparatus, such as the second device being a NWDAF or an AF.

[0210] In the embodiments of the present application, the first VFL task can also be referred to as or replaced by: a first VFL, a first VFL function, a first VFL process, or a first VFL activity, and the like, without limitation. For example, the first VFL task can include one or more of the following: network performance analysis, Artificial Intelligence (AI) model training, face recognition, or cross-institutional medical data analysis and disease prediction, and the like.

[0211] The candidate participating entity supporting the first VFL task can be understood as a candidate participating entity having the capability of performing the first VFL, or a candidate participating entity supporting the execution of the first VFL task. For example, the candidate participating entity can be a network element or a device, such as the candidate participating entity can include a NWDAF and / or an AF.

[0212] In the embodiments of the present application, the candidate participating entity determined from the candidate participating entities participating in the first VFL task can be referred to as a target participating entity, which can be understood as a candidate participating entity having the capability of participating in the first VFL task, or a candidate participating entity authorized to participate in the first VFL task, or an entity having the right to participate in the first VFL task. For the execution of the first VFL task, multiple target participating entities are required to participate, and the types of different target participating entities can be the same, different, partially the same, or partially different, which is not limited. The target participating entity can be a candidate participating entity having the capability of participating in the first VFL task selected from the candidate participating entity set supporting the first VFL task, so the target participating entity participating in the first VFL task can form a target participating entity set, which can be a subset or the whole set of the candidate participating entity set supporting the first VFL task. For example, if the candidate participating entity set includes candidate participating entities 1-6, the finally determined target participating entity set can include candidate participating entity 1, candidate participating entity 2, and candidate participating entity 5.

[0213] For the candidate participating entity supporting the first VFL, it is registered in the network, and the corresponding configuration information can be stored in the network storage network element, for example, the network storage network element can be a NRF or a NEF. Therefore, the second device can trigger the discovery process of the network storage network element to obtain one or more candidate participating entities supporting the first VFL task. It should be understood that the network storage network element can also be referred to as a candidate participating entity discovery network element, a discovery network element, etc., which is not limited in the embodiments of the present application.

[0214] In a possible implementation, the second device can send a third request message to the network storage network element, and correspondingly, the network storage network element receives the third request message from the second device. The third request message is used to request to discover the candidate participating entity supporting the first VFL task, and the third request message can include one or more of the following: an analysis identifier corresponding to the first VFL task, an interoperation identifier corresponding to the analysis identifier, a vertical federation group identifier, a requested role, etc.

[0215] Further, the network storage network element can determine candidate participating entities supporting the first VFL task according to the configuration information of the VFL of the entity stored by the network storage network element and the third request message, and send a response message corresponding to the third request message to the second device, wherein the response message includes information of the determined candidate participating entities, such as an identifier of the candidate participating entity or a list of identifiers of the candidate participating entities.

[0216] Optionally, the third request message can be a discovery request message in S204 or S304, such as a Nnrf_NFDiscovery_Request message, and the response message corresponding to the third request message can be a discovery response message in S206 or S306, such as a Nnrf_NFDiscovery_Response message.

[0217] Therefore, the second device can obtain a candidate participating entity set supporting the first VFL task according to the received response message corresponding to the third request message, and the candidate participating entity set includes at least one candidate participating entity. For example, the candidate participating entity set includes M candidate participating entities, and M is a positive integer greater than 1.

[0218] Further, the second device can request the corresponding candidate participating entity to obtain information of a third sample corresponding to the candidate participating entity as a first role. The first role can include a master participant and / or a slave participant. The candidate participating entity as a master participant supports providing a label in the first VFL task, and the label is used for the first VFL task, such as training of a model corresponding to the first VFL task and / or model evaluation; the candidate participating entity as a slave participant supports providing training data in the first VFL task, and the training data is used for the first VFL task as input data of a model corresponding to the first VFL task, for training and / or inference of the model. Optionally, the candidate participating entity as a master participant can also provide the label and the training data.

[0219] When the first role includes a master participant and a slave participant, the information of the third sample corresponding to the candidate participating entity as the first role can include information of a sample corresponding to the candidate participating entity as a master participant and information of a sample corresponding to the candidate participating entity as a slave participant, which is equivalent to data including all samples related to the first VFL task that can be provided by the candidate participating entity.

[0220] It should be understood that in the embodiments of the present application, the training data is input data of a model corresponding to the first VFL task or model training, and the label is data for comparing output data of the model training corresponding to the first VFL task, for converging the model. For example, the training data and the label can constitute one or more groups of data such as (x i , y i) sample data, i is the serial number of the sample data, x i is the training data, y i For label data, during model training, the y i Can be used with model input x i The output data at the time of the training is compared, and based on the comparison results, the trainable parameters in the model are adjusted through methods such as back propagation and gradient descent to obtain a model for inference.

[0221] The first role may refer to a role that the candidate participating entity can support in the first VFL task. This can be understood as the role that the candidate participating entity can assume / support in the first VFL task. The first role may also refer to a role that the second device indicates the candidate participating entity supports in the first VFL task. This can be understood as the role that the second device expects or requests the candidate participating entity to assume / support in the first VFL task. It should be understood that when the first role is the role that the second device expects or requests the candidate participating entity to assume / support in the first VFL task, the first role refers to a master participant or a slave participant.

[0222] In the embodiments of the present application, a role can be understood as a logical role in the first VFL task, such as a master participant, a slave participant, or a coordinator. A supported role can also be understood as a capability possessed, or a capability corresponding to a role, including the capability to provide labels in the first VFL task and / or the capability to provide training data in the first VFL task.

[0223] In one possible implementation, the second device may send a first request message to the candidate participating entity, and correspondingly, the first request message receives the first request message from the second device. The first request message is used to request information about the sample corresponding to the first role of the candidate participating entity, and the first request message may include one or more of the following: an identifier of the candidate participating entity, information about the first role, and instruction information for instructing to obtain information about the sample corresponding to the first role.

[0224] Optionally, the first request message may also be used to request the candidate participating entity to agree to participate in the first VFL task. In this case, the first request message may also include one or more of the following: an analysis identifier of the first VFL task, an interoperability indicator, and other information related to the first VFL task. In other words, while requesting the candidate participating entity to agree to participate in the first VFL task, the second device also requests information about the sample corresponding to the candidate participating entity as the first role.

[0225] Exemplarily, the second device can send the first request message to each of the M candidate participating entities respectively, and if the first role is indicated by the second device, the second device can indicate different first roles for different candidate participating entities.

[0226] Correspondingly, after receiving the first request message, each candidate participating entity can confirm whether to participate in the first VFL task, and if agreeing to participate in the first VFL task, the candidate participating entity can send information of a sample corresponding to the first role to the second device, such as the third sample corresponding to the first role. The information of the third sample corresponding to the first role of the candidate participating entity can be carried in the response message corresponding to the first request message and sent, such as the first response message.

[0227] Optionally, the first response message can also include a role supported by the candidate participating entity or a role agreed to be supported by the candidate participating entity.

[0228] Optionally, the first request message can be the VFL accurate request message in S207 or S307, and the first response message can be the VFL preparation response message in S209 or S309.

[0229] It should be understood that not all of the M candidate participating entities agree to participate in the first VFL task or will feed back the information of the sample corresponding to the first role. After receiving the first request message, some candidate participating entities can refuse to participate in the first VFL task after confirmation. The candidate participating entities that cannot participate in the first VFL task or cannot feed back the sample information can still feed back the first response message. At this time, the first response message can include the reason why the candidate participating entity cannot participate in the first VFL task or cannot feed back the sample information, such as the candidate participating entity does not have the information of the sample corresponding to the indicated first role, or other reasons.

[0230] Exemplarily, after the second device sends the first request message to the M candidate participating entities, the second device receives the information of the third sample corresponding to the first role fed back by N candidate participating entities respectively. At this time, N is a positive integer less than or equal to M.

[0231] Therefore, the second device can obtain the information of the third sample corresponding to the first role of the candidate participating entity agreeing to participate in the first VFL task according to the first response message fed back by the candidate participating entity. It should be understood that based on the different types of the first VFL task, the types and data of the samples provided by different candidate participating entities are also different.

[0232] The third sample corresponding to the candidate participating entity as the first role can be understood as a sample owned or supported by the candidate participant as the first role. Different roles provide different sample data for the same sample. The third sample can include one or more samples, and each sample corresponds to a sample identifier. Therefore, the information of the third sample can include the sample identifier and sample data, and the sample data can include a label and / or training data. The sample identifier can be configured by the candidate participating entity or can be an identifier of the sample itself. For example, the sample is a UE, and the sample identifier can be a permanent identifier SUPI of the UE or a local identifier configured by the candidate participating entity and having a mapping relationship with the SUPI.

[0233] If the first role is a main participant, the information of the third sample can include the identifiers of the samples supported by the candidate participant as the main participant and the labels corresponding to the samples or the labels corresponding to the samples and the training data. If the first role is a secondary participant, the information of the third sample can include the identifiers of the samples supported by the candidate participant as the secondary participant and the training data corresponding to the samples. If the first role includes a main participant and a secondary participant, the information of the third sample can include the identifiers of the samples supported by the candidate participant as the main participant and the labels corresponding to the samples, and the identifiers of the samples supported by the candidate participant as the secondary participant and the training data corresponding to the samples. The information of the third sample is equivalent to the information of all the samples supported by the candidate participant.

[0234] Optionally, the information of the third sample can further include one or more of the number of samples supported by the first role, the number of supported features, and the like. The number of samples supported by the first role can be understood as the number of samples included in the third sample, and the number of supported features can refer to the number of features extracted from the training data.

[0235] For example, the sample required by the first VFL task is a UE, N candidate participating entities own a plurality of UE labels and / or training data, the candidate participating entity is an AF, and N=2. As shown in Table 2 below, AF2 and AF3 can both support the roles of a main participant and a secondary participant in the first VFL task. B1-B4 are identifiers of UE samples allocated by AF2, C1-C4 are identifiers of UE samples allocated by AF3, B1 corresponds to the same UE sample (e.g., UE1) as C1, B2 corresponds to the same UE sample (e.g., UE2) as C2, B3 corresponds to the same UE sample (e.g., UE3) as C3, and B4 corresponds to the same UE sample (e.g., UE3) as C4. The identifier of the same UE sample has a corresponding relationship with the identifier of the UE.

[0236] Table 2

[0237] As can be seen from Table 2, AF2 supports the samples of UEs 1-4, and AF3 supports the samples of UEs 1-3 but does not support the sample of UE 4, that is, AF2 and AF3 support three same samples, but the sample data (including labels and / or training data) corresponding to the same samples are different.

[0238] In the case where the first role is a role expected by the second device for the candidate participating entity to support in the first VFL task, if the first role expected by the second device for AF2 to support in the first VFL task is a follower, and the first role expected by the second device for AF3 to support in the first VFL task is a leader, the information of the third samples corresponding to AF2 as the first role can include B1-B4 and the training data corresponding to B1-B4 respectively, and the information of the third samples corresponding to AF3 as the first role can include C1-C3 and the labels corresponding to C1-C3 respectively. Optionally, the information of the third samples corresponding to AF3 as the first role can further include the training data corresponding to C2 and C3 respectively.

[0239] S502, the second device determines, according to the information of the second samples supported by the first device and the information of the third samples corresponding to the candidate participating entity as the first role, the information of the first samples corresponding to the candidate participating entity as the first role.

[0240] In the embodiments of the present application, the first device is a device initiating the first VFL task, and the first device is a participating entity of the first VFL task, that is, the first device is a target participating entity participating in the first VFL task, and is also a task initiator of the first VFL task. For example, the first device can be a NWDAF or an AF.

[0241] The information of the second samples supported by the first device can be sent by the first device to the second device when the first device initiates the first VFL task. For example, the second device receives a second request message from the first device, and the second request message is a message used by the first device to request initiation of the first VFL task, for example, the second request message is used to request execution of the first VFL task, request the first VFL service, request the second device to provide a vertical federated learning service, or request the second device to obtain a candidate participating entity participating in the first VFL, and the like. Therefore, the information of the second samples supported by the first device can be carried in the second request message and sent.

[0242] Since the first device participates in the first VFL task, the second request message can also be used to indicate that the first device participates in the first VFL task, and the second request message can include information for indicating that the first device participates in the first VFL task. Optionally, the information for indicating that the first device participates in the first VFL task can be used not only to indicate that the first device participates in the first VFL task, but also to indicate the role played by the first device in the VFL task. Optionally, the information of the second samples supported by the first device can be included in the information for indicating that the first device participates in the first VFL task, or can be sent as an independent information element.

[0243] It should be understood that the second request message can also include one or more of the information related to the first VFL task, such as the analysis identifier corresponding to the first VFL task, the interoperability indication, and the like.

[0244] The process of obtaining the sample information of the candidate participating entity by the second device in S501 can be triggered by the second request message, in other words, the second device obtains the information of the third sample corresponding to the candidate participating entity as the first role according to the second request message. The information carried in the third request message obtained by the second device from the candidate participating entity can be determined according to the second request message.

[0245] After the second device obtains the information of the third sample corresponding to the candidate participating entity as the first role, the second device can filter or screen the information of the third sample according to the information of the second sample supported by the first device, and screen out the information of the sample not supported by the first device from the information of the third sample, to obtain the information of the first sample corresponding to the candidate participating entity as the first role. The first sample is the intersection of the second sample and the third sample.

[0246] Taking the first device AF1 as an example, the information of the second sample supported by the first device is shown in Table 3 below, A1-A4 are the identifiers of the samples allocated by AF1 to UE1-UE4 respectively, AF1 supports the samples of UE1-UE3, and does not support the sample of UE4, which is the same as the samples supported by AF2 and AF3. The sample data is different.

[0247] Table 3

[0248] Continuing to refer to the above example, the second device respectively obtains the information of the third sample corresponding to AF2 as the first role and the information of the third sample corresponding to AF3 as the first role. Since the second sample supported by AF1 includes the samples of UE1-UE3, the second device can remove the information of the sample not supported by AF1 in the information of the third sample of AF2 according to the correspondence between the sample identifier and the UE identifier, i.e., the training data with the sample identifier B4, to obtain the information of the fourth sample corresponding to AF2 as the follower. The information of the fourth sample corresponding to AF2 as the follower includes the sample identifiers B1-B3 and the training data corresponding to B1-B3 respectively. Similarly, the second device can remove the information of the sample not supported by AF1 in the information of the third sample of AF3. In the above example, since the third sample of AF3 is the same as the second sample supported by AF1, the information of the fourth sample corresponding to AF3 as the main role is the information of the third sample corresponding to AF3 as the main role, i.e., including the sample identifiers C1-C3 and the labels corresponding to C1-C3 respectively.

[0249] Further, in order not to disclose the information of the candidate participating entity which is not much related to the first VFL task to the task initiator, the second device can assign an identifier to each same sample of the candidate participating entity supported by the first device, and establish a mapping relationship between the assigned identifier and the sample identifier assigned by the first device to each sample, such as mapping A1, B1 and C1 to VFL UE1, mapping A2, B2 and C2 to VFL UE2, and mapping A3, B3 and C3 to VFL UE3, as shown in Table 4 below.

[0250] Table 4

[0251] Therefore, the second device can replace the sample identifier in the information of the fourth sample corresponding to AF2 as the follower with the sample identifier assigned by the second device to obtain the information of the first sample corresponding to AF2 as the follower. Similarly, the second device can replace the sample identifier in the information of the fourth sample corresponding to AF3 as the main role with the sample identifier assigned by the second device to obtain the information of the first sample corresponding to AF3 as the main role.

[0252] In a possible scenario, in the case that the candidate participating entity is an AF, since the AF needs to be authenticated through the NEF, the information of the third samples corresponding to the first role fed back by the AF to the second device through the NEF can be mapped by the NEF into a unified identifier for the same samples in the information of the third samples corresponding to the first role of each AF, the same identifier mapping as that of the second device is performed, and then the sample information after the same sample identifier mapping is sent by the NEF to the second device, so that the second device can filter the third sample information of each AF after the identifier mapping according to the second sample information supported by the first device, remove the sample information of the first device that is not supported in the third sample information of each AF after the identifier mapping, and obtain the first sample information of each AF corresponding to the first role.

[0253] S503, the second device sends the first sample information of the candidate participating entity corresponding to the first role to the first device. Correspondingly, the first device receives the first sample information of the candidate participating entity corresponding to the first role from the second device.

[0254] The second device sends the first sample information of each candidate participating entity after the sample filtering, sample identifier mapping and other processing to the first device, so that the first device determines the candidate participating entity participating in the first VFL task, that is, the target participating entity, and the role in the first VFL task. For details, see S504 below.

[0255] Optionally, the first sample information of the candidate participating entity corresponding to the first role can be sent in the response message corresponding to the second request message in S502 above, or in other defined or newly added messages, such as a newly added federated learning member confirmation request or a VFL member confirmation request, which is not limited.

[0256] S504, the first device determines the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task according to the first sample information of the candidate participating entity corresponding to the first role.

[0257] After the first device receives the first sample information of each candidate participating entity corresponding to the first role, the first device can select the candidate participating entity as the target participating entity and / or the role of the candidate participating entity as the target participating entity in the first VFL task according to one or more of the sample quantity, feature quantity and label quantity contained in the first sample information of each candidate participating entity corresponding to the first role.

[0258] It should be understood that the first device can determine the total number of target participating entities participating in the first VFL task according to the requirements of the first VFL task, and select the target participating entities based on the sample information of the feedback candidate participating entities, and / or the roles of the target participating entities in the first VFL task.

[0259] In a possible implementation, the first device can select the candidate participating entities as the target participating entities according to one or more of sample quantity priority, feature quantity priority, or label quantity priority. The sample quantity priority can refer to selecting the candidate participating entities whose sample quantity satisfies a threshold as the target participating entities, the feature quantity priority can refer to selecting the candidate participating entities whose feature quantity corresponding to the samples satisfies a threshold as the target participating entities, and the label quantity priority can refer to selecting the candidate participating entities whose label quantity corresponding to the samples satisfies a threshold as the target participating entities.

[0260] For example, the first device is AF1, which supports information of five samples of UEs 1-5, and the role in the first VFL task is a main participant. The obtained candidate participating entities include AF2-AF6, AF2 supports information of four samples of UEs 1-4 and the first role is a slave participant, AF3 supports information of three samples of UEs 1-3 and the first role is a main participant, AF4 supports information of four samples of UEs 1-3 and UE5 and the first role is a slave participant, AF5 supports information of five samples of UEs 1-5 and the first role is a slave participant, and AF6 supports information of two samples of UEs 1 and 2 and the first role is a main participant. Therefore, the first device AF1 can first retain candidate AFs with more sample quantities according to the sample quantities of the candidate AFs, such as AF2-AF5. If AF1 determines that the first VFL task requires four participants, since AF1 itself is a target participating entity, and it is a main participant, AF1 can further select three AFs from AF2-AF5 as target participating entities according to the feature quantity priority and label quantity priority strategies.

[0261] For example, since AF2 supports providing training data corresponding to sample UEs 1-4, AF3 supports providing labels corresponding to sample UEs 1-3, AF4 supports providing training data corresponding to sample UEs 1-3 and UE5, and AF3 supports providing training data corresponding to sample UEs 1-5, AF1, AF2, and AF3 can provide labels as master participants in the first VFL task, and AF1, AF2, AF4, and AF5 can provide training data as slave participants in the first VFL task. If the number of labels corresponding to each of the three samples provided by AF3 does not meet the threshold requirement, and the number of features in the training data of the same sample provided by AF2, AF4, and AF5 is comparable, AF1 can determine AF2, AF4, and AF5 as target participant entities, and determine the role of AF2 in the first VFL as a slave participant, the role of AF4 in the first VFL as a slave participant, and the role of AF5 in the first VFL as a slave participant according to the number of features of the sample.

[0262] Optionally, the first apparatus can also select the candidate participant entities as target participant entities and / or the roles of the candidate participant entities as target participant entities in the first VFL task in combination with at least one of the following: the weight of the features of the corresponding sample of the role supported by the candidate participant entity, the membership weight corresponding to the candidate participant entity, the membership corresponding to the candidate participant entity, the role supported by the candidate participant entity in the first VFL task, the computing capability of the candidate participant entity, or the hardware capability of the candidate participant entity.

[0263] wherein the member can refer to a participant who joins the VFL and is capable of participating in the VFL task or sharing resources with each other for the purpose of achieving common results, the membership corresponding to the candidate participant entity can be a natural person, a company, an organization, a network element entity (such as an application function or a service), or a network (such as a PLMN or a CN), etc., or can be a combination of any of the above members.

[0264] In a possible implementation, the first device can configure mandatory members and optional members participating in the first VFL task according to requirements of the first VFL task, so that the first device can determine one or more groups of candidate participating entities meeting the member requirements to participate in the first VFL task according to the member identities of the obtained candidate participating entities, and then determine the candidate participating entity as the target participating entity and / or the role of the candidate participating entity as the target participating entity in the first VFL task according to the information of the first sample of each candidate participating entity as the first role. The mandatory member refers to a member that must participate in the first VFL task, and the optional member refers to a member that can optionally participate in the first VFL task, that is, the target participating entity in the first VFL task can not include a participating entity with an optional member identity, but must include a participating entity with a mandatory member identity.

[0265] It should be understood that whether a participating entity with an optional member identity is stored in the first VFL task can be determined according to one or more of the number of target participating entities, the sample size, the feature quantity, or the label quantity required by the first VFL task.

[0266] For example, the first device AF1 determines that the mandatory members participating in the first VFL task include member 1 and member 2, and the optional members include member 3 and member 4, the member identity of AF1 is member 1, the first device obtains the information of the first sample of candidate participating entities AF2 to AF6 as the first role, if the member identity of AF2 is member 2, the member identity of AF3 is member 1, the member identity of AF4 is member 3, the member identity of AF5 is member 4, and the member identity of AF6 is member 5, the first device can determine that AF2 and AF3 are entities corresponding to the mandatory members according to the member identities of the candidate AFs, and AF4 and AF5 are entities corresponding to the optional members, and AF6 is neither an entity corresponding to the mandatory members nor an entity corresponding to the optional members, so the candidate participating entity combination meeting the member requirements can include {AF2, AF3}, {AF2, AF3, AF4}, {AF2, AF3, AF5}, and {AF2, AF3, AF4, AF5}.

[0267] Further, the AF1 can select a set of candidate participating entity combinations satisfying the first VFL task requirement according to one or more of the sample quantity, feature quantity, or label quantity of each candidate AF in each candidate participating entity combination, such as {AF2, AF3, AF4, AF5}, take the candidate participating entities in the candidate participating entity combination as the target participating entities, and further determine the roles of each candidate participating entity as the target participating entity in the first VFL task according to the first role of each candidate participating entity and the sample information. It should be understood that the finally determined target participating entity participating in the first VFL task is the one containing the first device AF1.

[0268] In another possible implementation, the first device can also assign a corresponding membership weight to each candidate participating entity of different membership, calculate a candidate participating entity combination with a higher weight value according to the obtained membership weight of each candidate participating entity, and further determine the candidate participating entity as the target participating entity and / or the role of the candidate participating entity as the target participating entity in the first VFL from the candidate participating entity combination in combination with the information of the first sample corresponding to each candidate participating entity as the first role, such as the sample quantity, feature quantity, and the like.

[0269] In some embodiments, the membership weight can be pre-configured locally in the first device, or can be obtained by the first device according to historical VFL task data, and is not limited in this regard.

[0270] Optionally, since the first device also serves as the target participating entity of the first VFL, the membership weight corresponding to the first device can also be involved in the calculation of the weight value.

[0271] For example, the first device AF1 obtains information of candidate participating entities AF2-AF6 as samples of the first role, the member identity corresponding to AF2 is member 2, the member identity corresponding to AF3 is member 1, the member identity corresponding to AF4 is member 3, the member identity corresponding to AF5 is member 4, and the member identity corresponding to AF6 is member 5. The member identity weights corresponding to members 1-5 assigned by the first device are w1, w2, w3, w4, and w5 in turn. Thus, the first device can calculate a combination of the member identity weights of each candidate AF to obtain a candidate participating entity combination with a higher weight value, such as W(AF2, AF3, AF4) = w2 + w1 + w3 = 0.8, W(AF2, AF4, AF5) = w2 + w3 + w4 = 0.6, …, W(AF2, AF3, AF5, AF6) = w2 + w1 + w4 + w5 = 0.7. If the weight value of W(AF2, AF3, AF4) is the highest, the first device AF1 can further determine whether AF2, AF3, and AF4 meet the sample requirements, feature requirements, and the like of the first VFL task according to the sample information of AF2, AF3, and AF4, so as to determine whether they can all be target participating entities. If they all meet the requirements or some of them meet the requirements, the AFs that meet the requirements can be determined as target participating entities, and the roles of the candidate participating entities that are determined as target participating entities in the first VFL task can also be determined according to the first roles corresponding to the sample information of the candidate participating entities.

[0272] In another possible implementation, the first device can configure a feature list according to the feature requirements of the samples required to complete the first VFL task, and select candidate participating entities having all or part of the features in the feature list from the sample information of the candidate participating entities obtained as the first role corresponding to the sample information, as target participating entities. The roles of the candidate participating entities that are determined as target participating entities in the first VFL task can also be determined according to the first roles corresponding to the sample information of the candidate participating entities.

[0273] In another possible implementation, the first device can also configure feature weights for features required to complete the first VFL task, different features correspond to different feature weights, and the optimal feature combination is calculated by using the feature weights, so that candidate participating entities having the feature combination can be selected from the sample information of the candidate participating entities obtained as the first role corresponding to the sample information, as target participating entities. The roles of the candidate participating entities that are determined as target participating entities in the first VFL task can also be determined according to the first roles corresponding to the sample information of the candidate participating entities.

[0274] For example, the first device determines that the features of the samples required for the first VFL task are {f1, f2, f3, f4, f5, f6}, and the corresponding assigned feature weights are w1, w2, w3, w4, w5, and w6 in sequence. If the first device calculates a more optimal feature combination {f1, f2, f3} based on the weights, the first device can select a candidate participating entity with the feature combination {f1, f2, f3} from the obtained multiple candidate participating entities as the target participating entity, and can also determine the role of the candidate participating entity as the target participating entity in the first VFL task according to the first role corresponding to the sample information of the candidate participating entity.

[0275] Optionally, the feature weights can be pre-configured locally in the first device, or can be obtained by the first device according to historical VFL task data, and the present application does not make any limitation in this regard.

[0276] The above examples show several processes in which the first device determines the candidate participating entity as the target participating entity participating in the first VFL task and / or the role of the candidate participating entity as the target participating entity in the first VFL task according to the information of the first sample corresponding to the candidate participating entity as the first role. In addition to considering the identity information and sample information of the candidate participating entity, the first device can also consider the hardware or computing capability of the candidate participating entity or other factors to determine the target participating entity and / or the corresponding role, and the present application does not make any limitation in this regard.

[0277] After the first device determines the target participating entity participating in the first VFL task and / or the role of the target participating entity in the first VFL task based on the above processes, the first device can inform the second device acting as the task coordinator, so that the second device performs sample alignment on the samples provided by the target participating entity to trigger execution of the first VFL task.

[0278] S505, the first device sends information of the target participating entity participating in the first VFL task and / or information of the role of the target participating entity in the first VFL task to the second device. Correspondingly, the second device receives the information of the target participating entity participating in the first VFL task and / or the information of the role of the target participating entity in the first VFL task from the first device.

[0279] The information of the target participating entity participating in the first VFL task can include the identification or identification list of the target participating entity, and the information of the role of the target participating entity in the first VFL task can include the role of each target participating entity in the first VFL task, such as a main participating entity providing labels or providing labels and training data and / or a slave participating entity providing training data.

[0280] Thus, the second device can determine which target participating entities participate in the first VFL task and / or what roles the target participating entities play in the first VFL task according to the obtained information of the target participating entities and / or the information of the roles of the target participating entities in the first VFL task, thereby performing sample alignment on sample data provided by each target participating entity as a corresponding role, and sending the aligned sample data to each target participating entity to perform model training and / or inference corresponding to the first VFL task.

[0281] Optionally, the first device can also send information of samples of the target participating entities for the first VFL task to the second device, which can include sample data after sample alignment on sample data provided by each target participating entity as a corresponding role, so as to send the sample data to other target participating entities to perform model training and / or inference corresponding to the first VFL task. In this case, the second device can not perform sample alignment processing.

[0282] Optionally, the information of the target participating entities participating in the first VFL task can include the aligned sample data.

[0283] It should be understood that in the case where the first device determines one of the target participating entities participating in the first VFL task or the roles of the target participating entities in the first VFL task, the other one can be determined by the second device based on the information fed back by the first device.

[0284] Based on the communication method shown in FIG. 5, the first device as a task initiator and a target participating entity obtains information of samples of candidate participating entities in a first role from the second device as a task coordinator, and the information of the samples in the first role is filtered by the second device using samples supported by the first device, so that the candidate participating entities as target participating entities and / or the roles of the candidate participating entities as target participating entities in the first VFL task can be determined according to the information of the samples in the first role.

[0285] The communication method shown in FIG. 5 will be described in detail below in combination with a 5G system. Taking an example in which the first device and the candidate participating entities are both AFs, and the second device is NWDAF, wherein the first device is AF1, and the candidate participating entities include AF2-AFM.

[0286] As shown in FIG. 6, the communication method includes:

[0287] S601, AF1-AFM and NWDAF are registered to NRF.

[0288] The AF1-AFM and the NWDAF register the VFL capability supported by them to the NRF, wherein the VFL capability can include an analysis ID, a joined vertical federated learning group (such as an AF ID, a Vendor ID, or a UE ID), a supported role attribute (such as a participant (whether a label and / or training data can be provided) or a coordinator), and the like. For details, refer to the related description in the existing implementation mode, such as S201-S203 or S301-S303, which will not be described herein.

[0289] S602, the AF1 sends a VFL preparation request message #1 to the NWDAF. Correspondingly, the NWDAF receives the VFL preparation request message #1 from the AF1.

[0290] The VFL preparation request message #1 is used to request the NWDAF to provide a vertical federated learning service, such as requesting the NWDAF to obtain a candidate AF participating in the first VFL task. The VFL preparation request message #1 can include at least one of the following: an analysis ID corresponding to the first VFL task, interoperability information corresponding to the first VFL task, a joint ML training ID corresponding to the first VFL task, indication information of the AF1 participating in the first VFL task, and a role of the AF1 participating in the first VFL task, and the like.

[0291] Since the AF1 not only acts as a task initiator or a task requester of the first VFL task, but also acts as a participant of the first VFL task, the AF1 can also indicate the AF1 participating in the first VFL task through the VFL preparation request message #1. At this time, the VFL preparation request message #1 can also include information indicating the AF1 participating in the first VFL task, such as a role of the AF1 in the first VFL task, information of a sample #2 supported by the AF1, and the like.

[0292] The VFL preparation request message #1 can correspond to the second request message in S502, and the specific description can be referred to the description of the second request message, which will not be described herein.

[0293] S603, the NWDAF sends a discovery request message to the NRF. Correspondingly, the NRF receives the discovery request message from the NWDAF.

[0294] The discovery request message is used to request to discover a candidate AF supporting the first VFL task, and the discovery request message corresponds to the third request message in S501. For details, refer to the description of the third request message, which will not be described herein.

[0295] S604, the NRF sends a discovery response message to the NWDAF. Correspondingly, the NWDAF receives the discovery response message from the NRF.

[0296] The discovery response message includes information of candidate AFs supporting the first VFL task, such as an identifier or an identifier list of the candidate AFs. For example, the candidate AFs supporting the first VFL task include AF2-AF6, and the discovery response message includes the AF identifiers corresponding to AF2-AF6 respectively.

[0297] The discovery response message corresponds to the response message corresponding to the third request message in S501. For details, refer to the description of the response message corresponding to the third request message, which will not be repeated here.

[0298] S605, the NWDAF sends a VFL preparation request message #2 to the candidate AF. Correspondingly, the candidate AF receives the VFL preparation request message #2 from the NWDAF.

[0299] The VFL preparation request message #2 is used to request information of samples corresponding to the first role of the candidate AF. The first role of each candidate AF can refer to a role that the candidate AF can support in the first VFL task, or a role that the NWDAF expects the candidate AF to assume in the first VFL task.

[0300] The VFL preparation request message #2 can include the identifier of the candidate AF, the information of the first role, and the information indicating the samples corresponding to the first role of the candidate AF. The VFL preparation request message #2 corresponds to the first request message in S501. For details, refer to the description of the first request message, which will not be repeated here.

[0301] S606, the candidate AF sends a VFL preparation response message #2 to the NWDAF. Correspondingly, the NWDAF receives the VFL preparation response message #2 from the candidate AF.

[0302] The VFL preparation response message #2 includes information of samples #1 corresponding to the first role of the candidate AF, which can include a sample identifier stored by the candidate AF and sample data corresponding to the first role.

[0303] The information of the samples #1 corresponding to the first role of the candidate AF corresponds to the information of the third samples corresponding to the first role of the candidate participating entity. For details, refer to the description of the information of the third samples corresponding to the first role of the candidate participating entity, which will not be repeated here.

[0304] S607, the NWDAF determines information of samples #3 corresponding to the first role of the candidate AF according to the information of the samples #1 corresponding to the first role of the candidate AF and the information of the samples #2 supported by the AF1.

[0305] After the NWDAF obtains the information of the sample #1 corresponding to each candidate AF as the first role, the NWDAF can filter the information of the sample #1 corresponding to each candidate AF as the first role by using the information of the sample #2 supported by the AF1, filter out the information of the sample that is not supported by the AF1 in the information of the sample #1 corresponding to each candidate AF as the first role, and thus obtain the information of the sample #3 corresponding to each candidate AF as the first role. The sample #3 is the intersection of the sample #1 and the sample #2.

[0306] In the case where the sample identifiers of the candidate AFs are not mapped by the NEF, after the NWDAF filters the sample information of each candidate AF, the NWDAF also needs to perform identifier mapping on the filtered sample information of each candidate AF, and map the same sample identifiers provided by the candidate AFs into one identifier, as shown in Table 4.

[0307] That is, the information of the sample #3 corresponding to each candidate AF as the first role includes the identifiers of the filtered and identifier-mapped samples, and the sample data corresponding to the identifiers as the first role.

[0308] The specific implementation process of S607 can be referred to the related description in S502. The sample #2 supported by the AF1 corresponds to the information of the second sample supported by the first device in S502, and the information of the sample #3 corresponding to each candidate AF as the first role corresponds to the information of the first sample corresponding to each candidate participating entity as the first role. Details are not described herein.

[0309] S608, the NWDAF sends a VFL information confirmation request message to the AF1. Correspondingly, the AF1 receives the VFL information confirmation request message from the NWDAF.

[0310] The VFL information confirmation request message is used to request the AF1 to determine the candidate AFs participating in the first VFL task and / or the roles of the candidate AFs in the first VFL task, and the VFL information confirmation request message includes the information of the sample #3 corresponding to each candidate AF as the first role.

[0311] The specific implementation process of S608 can be referred to the related description in S503, and details are not described herein.

[0312] S609, the AF1 determines the candidate AFs participating in the first VFL task and / or the roles of the candidate AFs in the first VFL task according to the information of the sample #3 corresponding to each candidate AF as the first role.

[0313] AF1 obtains information of each candidate AF as a sample #3 corresponding to the first role, and can select a candidate AF satisfying the requirement of the first VFL task as the target participating entity according to the number of samples, the number of features, the label data and the like provided by each candidate AF as the first role, and determine the role of the candidate AF as the target participating entity in the first VFL task based on the sample information provided by the candidate AF.

[0314] The specific implementation process of S609 can be referred to the related description in S504, which will not be repeated here.

[0315] S610, AF1 sends information of the candidate AF participating in the first VFL task and / or information of the role of the candidate AF in the first VFL task to the NWDAF.

[0316] The information of the candidate AF participating in the first VFL task includes the identifier of the candidate AF, and the information of the role of the candidate AF in the first VFL task includes the role of each candidate AF in the first VFL task.

[0317] The specific implementation process of S610 can be referred to the related description in S505, which will not be repeated here.

[0318] The above FIG. 5 and FIG. 6 describe the implementation process of how the first device as the task initiator and / or participant determines the target participating entity of the first VFL task and / or the corresponding role in the first VFL task. In addition, the embodiments of the present application also provide a communication method for determining the target participating entity of the first VFL task and / or the corresponding role in the first VFL task by the second device as the task coordinator. At this time, the first device can participate in the first VFL task as a candidate participating entity or not.

[0319] For example, FIG. 7 is a flowchart of another communication method provided by the embodiments of the present application, which still takes the communication between the first device and the second device in the communication system shown in FIG. 4 as an example.

[0320] As shown in FIG. 7, the communication method includes:

[0321] S701, the second device obtains information of a third sample corresponding to the first role of a candidate participating entity supporting the first VFL task.

[0322] In a possible implementation, the second device can send a first request message to the candidate participating entity, where the first request message is used to request information of the candidate participating entity as a sample corresponding to the first role, so as to obtain, according to a first response message sent by the candidate participating entity according to the first request message, information of the candidate participating entity as a third sample corresponding to the first role, that is, the first response message includes / carry information of the candidate participating entity as the third sample corresponding to the first role. For details, refer to the related description of the second device obtaining information of the candidate participating entity as the third sample corresponding to the first role in support of the first VFL task in S501 above, which is not repeated here.

[0323] The second device obtaining the candidate participating entity can be implemented by a discovery service process of a network storage network element. Since the second device is a task coordinator of the first VFL task, the execution of S701 can be triggered by the first device, for example, by the following second request message, which is not repeated here. Different from the above embodiment, in the embodiment shown in FIG. 7, the first device can be an initiator of the first VFL task, or can be another entity or device with VFL capability, or can be an entity or device without VFL capability, for example, a subscriber of a VFL service, which is not limited.

[0324] The specific implementation process of S701 is similar to that of S501, and for details, refer to the related description in S501 above, which is not repeated here.

[0325] In the case that the first device wants to participate in the first VFL task, it can be considered that the first device has the capability to support the first VFL task, and the first device can request to participate in the first VFL task as a candidate participating entity. At this time, the candidate participating entity supporting the first VFL task obtained by the second device includes the first device, in other words, the information of the candidate participating entity supporting the first VFL task as the third sample corresponding to the first role includes the information of the first device as the third sample corresponding to the first role. The first device as the candidate participating entity supporting the first VFL task can be discovered by the second device through a discovery service process of a network storage network element, or can be actively initiated by the first device.

[0326] In the case that the first device actively requests to participate in the first VFL task, the information of the first device as the third sample corresponding to the first role can be actively sent by the first device to the second device, which is different from obtaining the information of other candidate participating entities as the third sample corresponding to the first role. In a possible design, the information of the first device as the third sample corresponding to the first role can be carried in the request message sent by the first device to initiate the first VFL task.

[0327] Under the design, the first device can send a second request message to the second device, and correspondingly, the second device can receive the second request message from the first device. The second request message has at least one of the following functions: a function of requesting to perform the first VFL task, a function of requesting the first VFL service, a function of requesting the second device to provide the vertical federated learning service, or a function of requesting the second device to obtain the candidate participating entity participating in the first VFL, and the information of the third sample corresponding to the first role of the first device can be carried in the second request message.

[0328] Optionally, the second request message can also be used by the first device to request to participate in the first VFL task, to indicate that the first device wants to participate in the first VFL task. Optionally, the second request message can further include one or more of the following: information for indicating that the first device requests to participate in the first VFL task, and the information for indicating that the first device requests to participate in the first VFL task can include a first role of requesting to participate and / or an indication of requesting to participate. At this time, the information for indicating that the first device requests to participate in the first VFL task can be information for indicating that the first device requests to participate in the first VFL task in the first role.

[0329] Optionally, the first device can also indicate the request to participate in the first VFL task through other defined or newly added messages. Alternatively, the information for indicating that the first device requests to participate in the first VFL task can also be carried in other defined or newly added messages.

[0330] Optionally, the information of the third sample corresponding to the first role of the first device can be included in the information for indicating that the first device requests to participate in the first VFL task.

[0331] Optionally, the information for indicating that the first device requests to participate in the first VFL task can be used only to indicate that the first device requests to participate in the first VFL task, and the first role of requesting to participate and the information of the third sample corresponding to the first role of the first device can be sent as independent information elements, that is, the three can be indicated through different information elements. At this time, the second request message can include the information for indicating that the first device requests to participate in the first VFL task, the first role of requesting to participate, and the information of the third sample corresponding to the first role of the first device, or the three are carried in different messages for sending, and no limitation is made in this regard.

[0332] Therefore, the second device can know that the first device requests to participate in the first VFL task, so as to determine whether the first device can serve as the target participating entity according to the information of the third sample corresponding to the first role of the first device. Optionally, the second device can also feed back a response of whether it participates in the first VFL task to the first device.

[0333] For example, the candidate participating entities obtained by the second device are all AFs, including AF1-AF3, wherein the first role of AF1 is a slave participant, the first role of AF2 is a master participant, and the first role of AF3 includes a master participant and a slave participant. The information of the third samples corresponding to the first roles of AF1-AF3 is shown in Table 5 below. A1-A4 are identifiers of the UE samples allocated to AF1, B1-B4 are identifiers of the UE samples allocated to AF2, and C1-C4 are identifiers of the UE samples allocated to AF3. A1, B1, and C1 correspond to the same UE sample (e.g., UE1), A2, B2, and C2 correspond to the same UE sample (e.g., UE2), A3, B3, and C3 correspond to the same UE sample (e.g., UE3), A4, B4, and C4 correspond to the same UE sample (e.g., UE3), and the identifiers of the same UE samples have a corresponding relationship with the identifiers of the UEs.

[0334] Table 5

[0335] Therefore, the third samples of AF1 as a slave participant include sample identifiers A1-A3 and the training data corresponding to A1-A3, respectively. The third samples of AF2 as a master participant include sample identifiers B1-B4 and the labels corresponding to B1-B4, respectively. The third samples of AF3 as a master participant and a slave participant include sample identifiers C1-C4, the labels corresponding to C1-C4, and the training data corresponding to C1-C3, respectively.

[0336] In S702, the second device determines the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task according to the information of the third samples corresponding to the candidate participating entities as the first roles.

[0337] In a possible design, the second device can determine the candidate participating entity as the target participating entity of the first VFL task and / or the role of the candidate participating entity in the first VFL task according to the information of the third samples corresponding to the candidate participating entities as the first roles based on a selection policy of the target participating entity and / or the corresponding role configured locally. For details, refer to the description of the first device determining the candidate participating entity as the target participating entity of the first VFL task and / or the role of the candidate participating entity in the first VFL task according to the information of the first samples corresponding to the candidate participating entities as the first roles in S504.

[0338] Optionally, the second device can select the candidate participant entity as the target participant entity and / or the role of the candidate participant entity in the first VFL task in combination with at least one of the following: the weight of the feature of the corresponding sample supported by the candidate participant entity, the membership weight corresponding to the candidate participant entity, the membership corresponding to the candidate participant entity, the role supported by the candidate participant entity in the first VFL task, the computing capability of the candidate participant entity, or the hardware capability of the candidate participant entity, similar to the first device in S504.

[0339] Unlike S504, the weight of the feature of the corresponding sample supported by the candidate participant entity and the membership weight corresponding to the candidate participant entity can be pre-configured locally on the second device or configured by the first device for the second device, without limitation.

[0340] In another possible design, the second device can determine the selection strategy of the target participant entity and / or the corresponding role, which can be configured by the first device. In a possible implementation, the first device can send first information to the second device, and the second device can receive the first information from the first device. The first information is used to indicate the method of determining the target participant entity of the first VFL task and / or the role of the target participant entity in the first VFL task. Thus, the second device can determine the candidate participant entity as the target participant entity of the first VFL task and / or the role of the candidate participant entity in the first VFL task according to the information of the third sample corresponding to the first role of the candidate participant entity and the first information.

[0341] Optionally, the first information can be carried in the second request message or other defined or newly added messages, without limitation.

[0342] Optionally, the information used to indicate that the first device requests to participate in the first VFL task can be included in the first information, which can be considered as a method of determining the target participant entity of the first VFL task and / or the role of the target participant entity in the first VFL task.

[0343] The first information can be understood as a selection policy, a selection method, a determination policy or a determination method of the target participating entity and / or the corresponding role, and the first information can indicate one or more methods of determining the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task. The method can still be set in combination with one or more of the weights of the features of the samples corresponding to the roles supported by the candidate participating entity, the membership weight corresponding to the candidate participating entity, the membership corresponding to the candidate participating entity, the role supported by the candidate participating entity in the first VFL task, the computing capability of the candidate participating entity, or the hardware capability of the candidate participating entity.

[0344] For example, the first information can indicate at least one of the following methods:

[0345] Method 1: determining the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task according to the weights of the features of the samples corresponding to the roles supported by the candidate participating entity and the information of the third sample corresponding to the first role of the candidate participating entity, wherein the first information can include the configured feature weights and the corresponding weight algorithm;

[0346] Method 2: determining the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task according to the membership weight corresponding to the candidate participating entity and the information of the third sample corresponding to the first role of the candidate participating entity, wherein the first information can include the configured membership weight corresponding to different membership respectively and the corresponding weight algorithm;

[0347] Method 3: determining the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task according to the membership corresponding to the candidate participating entity and the information of the third sample corresponding to the first role of the candidate participating entity, wherein the first information can include the configured identity weight corresponding to different membership respectively, and the first information can include the mandatory members and the optional members participating in the first VFL task;

[0348] Method 4: determining the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task according to one or more of the sample quantity, the feature quantity or the label quantity in the information of the third sample corresponding to the first role of the candidate participating entity, wherein the first information can include one or more of the sample quantity threshold, the feature quantity threshold or the label quantity threshold to be met by each candidate participating entity.

[0349] The above exemplary shows four methods of determining target participating entities and / or corresponding roles, whether the second device determines the target participating entities and / or corresponding roles based on local policies or based on the first information. Specific examples can refer to the related description in S504 above, which will not be repeated here. At this time, the information of the candidate participating entities in S504 above as the first sample of the first role can be understood as the information of the candidate participating entities in S702 as the third sample of the first role.

[0350] Therefore, after determining the target participating entities participating in the first VFL task and the roles of the target participating entities in the first VFL task, the second device can perform sample alignment on the information of the samples of the target participating entities corresponding to the roles, and trigger the target participating entities to perform model training and / or inference corresponding to the first VFL task by using the aligned samples.

[0351] Continuing to refer to the example in S701 above, the second device determines that the target participating entities include AF1-AF3, and the roles of AF1, AF2, and AF3 in the first VFL task are follower, main participant, and main participant and follower, respectively. The second device can perform sample alignment processing. Since AF1 does not support the UE sample corresponding to A4, the sample data of each AF after sample alignment is shown in Table 6 below.

[0352] Table 6

[0353] Further, the second device can also map the identifiers of the same samples corresponding to each AF to uniform identifiers, and different samples correspond to different uniform identifiers to avoid information leakage. The uniform identifiers are configured by the second device. Exemplarily, the mapped identifiers are shown in Table 4 above. The specific implementation process can refer to the related description of identifier mapping in S502 above, which will not be repeated here.

[0354] Optionally, after determining the target participating entities and / or corresponding roles, the second device can inform the first device, such as performing S703 below.

[0355] S703. The second device sends, to the first device, information of the target participating entities participating in the first VFL task and / or information of the roles of the target participating entities in the first VFL task.

[0356] The related description of the information of the target participating entities participating in the first VFL task and the information of the roles of the target participating entities in the first VFL task can refer to the related description in S505 above, which will not be repeated here.

[0357] Optionally, if the first device participates in the first VFL task, the second device can further send, to the first device, information of samples of each target participating entity as a corresponding role for performing the first VFL task after sample alignment and sample identification mapping, so as to facilitate the first device to perform model training and / or inference corresponding to the first VFL task.

[0358] Based on the communication method shown in FIG. 7, the second device as the task coordinator determines the candidate participating entity as the target participating entity and / or the role of the candidate participating entity as the target participating entity in the first VFL task according to the information of the samples of the candidate participating entity as the first role.

[0359] The communication method shown in FIG. 7 is described in detail below in combination with a 5G system. Taking the first device as NWDAF1, the second device as NWDAF2, and the candidate participating entity as AF as an example, the candidate participating entity includes AF1 to AFM.

[0360] As shown in FIG. 8, the communication method includes the following steps.

[0361] S801, AF1 to AFM, NWDAF1, and NWDAF1 are registered to the NRF.

[0362] The specific implementation process of S801 can be referred to the related description in S601 described above, and will not be described here.

[0363] S802, NWDAF1 sends a VFL preparation request message #1 to NWDAF2. Correspondingly, NWDAF2 receives the VFL preparation request message #1 from NWDAF1.

[0364] The VFL preparation request message #1 is used to request NWDAF2 to provide a vertical federated learning service, such as requesting NWDAF2 to obtain candidate AFs participating in the first VFL task.

[0365] Optionally, in the case where NWDAF1 participates in the first VFL task, the VFL preparation request message is further used to indicate that NWDAF1 requests to participate in the first VFL task, and the VFL preparation request message #1 can include role information and / or sample data corresponding to the role of NWDAF1 participating in the first VFL task request, and other information related to the request to participate in the first VFL task.

[0366] Optionally, the VFL preparation request message #1 can be further used to indicate a policy for NWDAF2 to determine the target participating entity and / or the corresponding role, such as the VFL preparation request message #1 including first information, and the first information is used to indicate the target participating entity performing the first VFL task and / or the role of the target participating entity in the first VFL task.

[0367] The specific implementation process of S802 can refer to the related description in S701 or S602 described above, and details are not described herein.

[0368] S803, the NWDAF 2 sends a discovery request message to the NRF. Correspondingly, the NRF receives the discovery request message from the NWDAF 2.

[0369] S804, the NRF sends a discovery response message to the NWDAF 2. Correspondingly, the NWDAF 2 receives the discovery response message from the NRF.

[0370] S805, the NWDAF 2 sends a VFL preparation request message #2 to the candidate AF. Correspondingly, the candidate AF receives the VFL preparation request message #2 from the NWDAF 2.

[0371] S806, the candidate AF sends a VFL preparation response message #2 to the NWDAF 2. Correspondingly, the NWDAF 2 receives the VFL preparation response message #2 from the candidate AF.

[0372] The specific implementation process of S803-S806 can refer to the related description in S603-S606 or S701 described above, and details are not described herein.

[0373] S807, the NWDAF 2 determines the candidate AF participating in the first VFL task and / or the role of the candidate AF in the first VFL task according to the information of the sample #1 corresponding to the first role of the candidate AF.

[0374] The specific implementation process of S807 can refer to the related description in S504 or S609 or S702 described above, and details are not described herein.

[0375] S808, the NWDAF 2 sends information of the candidate AF participating in the first VFL task and / or information of the role of the candidate AF in the first VFL task to the NWDAF 1. Correspondingly, the NWDAF 1 receives the information of the candidate AF participating in the first VFL task and / or the information of the role of the candidate AF in the first VFL task from the NWDAF 2.

[0376] The specific implementation process in S808 can refer to the related description in S703 described above, and details are not described herein.

[0377] It should be understood that, in the above examples, the types of the first device and the second device can also be other NFs or AFs, which are not limited. Moreover, the names of the messages and information shown in the embodiments of the present application are not limited.

[0378] In the above various embodiments, the method and / or steps implemented by the first device can also be implemented by components (e.g., a processor, a chip, a chip system, a circuit, a logic module, or software) available to the first device; the method and / or steps implemented by the second device can also be implemented by components (e.g., a processor, a chip, a chip system, a circuit, a logic module, or software) available to the second device.

[0379] The above mainly introduces the solutions provided in the present application. Accordingly, the present application also provides a communication device for implementing various methods in the above method embodiments. The communication device can be the first device in the above method embodiments, or a device containing the first device, or a component available to the first device, such as a chip or a chip system. Alternatively, the communication device can be the second device in the above method embodiments, or a device containing the second device, or a component available to the second device, such as a chip or a chip system.

[0380] In some embodiments, the communication device contains hardware structures and / or software modules for performing respective functions in order to implement the above functions. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of the examples described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is implemented in hardware or computer software driven 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 the present application.

[0381] The embodiments of the present application can divide the functional modules of the communication device according to the above method embodiments, for example, each functional module can be divided according to each function, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware or software functional module. It should be noted that the division of modules in the embodiments of the present application is illustrative, and is only a logical functional division. Actual implementation can have another division manner.

[0382] Taking the communication device as the first device or the second device in the above method embodiments, Fig. 9 is a structural schematic diagram of a communication device provided in an embodiment of the present application. As shown in Fig. 9, the communication device 900 includes a processing module 901 and a communication module 902. The processing module 901 is configured to perform the processing functions of the first device or the second device in the above method embodiments. The communication module 902 is configured to perform the communication functions of the first device or the second device in the above method embodiments.

[0383] All the related content of each step involved in the method embodiments can be cited to the function description of the corresponding function module, and will not be repeated here.

[0384] In a possible design, the communication module 902 can include a receiving module and a sending module (not shown in FIG. 9). The sending module and the receiving module are respectively used to implement the sending function and the receiving function of the communication apparatus 900.

[0385] In a possible design, the communication apparatus 900 can further include a storage module (not shown in FIG. 9), which stores programs or instructions. When the processing module 901 executes the programs or instructions, the communication apparatus 900 can perform the functions of the first device or the second device in any of the methods shown in FIGS. 5-8.

[0386] In some embodiments, the processing module 901 involved in the communication apparatus 900 can be implemented by a processor or a processor-related circuit component, and can be a processor or a processing unit; the communication module 902 can be implemented by a transceiver or a transceiver-related circuit component, and can be a transceiver or a receiving unit.

[0387] Exemplarily, FIG. 10 is a structural schematic diagram of another communication apparatus provided by an embodiment of the present application. The communication apparatus can be the first device or the second device in the method embodiments, or can be a chip (system) or other components or assemblies that can be arranged in the first device or the second device. As shown in FIG. 10, the communication apparatus 1000 can include a processor 1001, a bus 1002, a communication interface 1003, and a memory 1004. The processor 1001, the memory 1004, and the communication interface 1003 communicate through the bus 1002. The communication apparatus 1000 can be the first device or the second device. It should be understood that the number of processors and memories in the communication apparatus 1000 is not limited in the present application.

[0388] The bus 1002 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one line is shown in FIG. 10, but it does not mean that there is only one bus or only one type of bus. The bus 1002 can include a path for transmitting information between various components (for example, the memory 1004, the processor 1001, and the communication interface 1003) of the communication apparatus 1000.

[0389] The processor 1001 can include any one or more of a CPU, a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP), among other processors.

[0390] The memory 1004 can include volatile memory (e.g., random access memory (RAM)), storage device (e.g., read only memory (ROM), floppy disk drive, hard disk drive, or solid state drive), or any other form of storage for storing data.

[0391] The communication interface 1003 uses a transceiver module such as, but not limited to, a network interface card, a transceiver, or the like, to enable communications between the communication apparatus 1000 and other devices or communication networks.

[0392] The memory 1004 stores executable program code that the processor 1001 executes to implement the functionality of the first device or the second device, respectively, as described in the foregoing method embodiments. That is, the memory 1004 has instructions stored thereon for performing the above-described communication method.

[0393] In yet another aspect, the embodiments of the present disclosure also provide a computer program product containing instructions, which, when executed on a communication apparatus, enable the communication apparatus to perform the method described in any of the embodiments.

[0394] In yet another aspect, the embodiments of the present disclosure also provide a computer readable storage medium. The computer readable storage medium stores computer programs or instructions, which, when executed on a communication apparatus, enable the communication apparatus to perform the method described in any of the embodiments.

[0395] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present application are generated. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (for example, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (for example, infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or data storage device such as one or more servers, data centers, etc. integrated with one or more media. The available media can be magnetic media (for example, floppy disk, hard disk, magnetic tape), optical media (for example, DVD), or semiconductor media (for example, solid state disk (SSD)) and the like.

[0396] Those skilled in the art can appreciate that the units and algorithm steps of the 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 the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0397] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0398] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the described device embodiments are merely schematic. The division of the units is merely logical function division. There can be other division manners in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0399] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0400] In addition, each functional unit in the various embodiments of the present application can be integrated into a processing unit, or each unit can exist physically, or two or more units can be integrated into one unit.

[0401] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product in essence or in the part that contributes to the prior art, or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or an access network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a ROM, a random access memory RAM, a magnetic disk or an optical disk, and various media that can store program codes.

[0402] Although the present application is described herein in conjunction with various embodiments, other variations of the disclosed embodiments can be understood and implemented by those skilled in the art with reference to the attached drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "one" does not exclude a plurality. A single processor or other unit can implement several functions listed in the claims. Some measures described in mutually different dependent claims can be combined and produce good results.

[0403] Although the present application has been described in connection with certain specific features and embodiments thereof, it is to be understood that it is provided as an exemplification of the application and is not intended to limit the scope of the application, which is defined in the claims. Various modifications and changes can be made thereto without departing from the spirit and scope of the application. Accordingly, it is intended that all such modifications and changes be included within the scope of the application as claimed. Obviously, many modifications and variations of the present application are possible in light of the above teachings. It is, therefore, to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A communication method, characterized in that: The method comprises: A first device receives information about a first sample corresponding to a first role as a candidate participating entity supporting a first vertical federated learning (VFL) task from a second device, where the first device is the device that initiates the first VFL task; The first device determines, based on information of the first sample corresponding to the first role of the candidate participating entity, a target participating entity for the first VFL task and / or a role of the target participating entity in the first VFL task; The first device sends information about the target participating entity participating in the first VFL task and / or information about the role of the target participating entity in the first VFL task to the second device.

2. The method according to claim 1, characterized in that The role of the target participating entity in the first VFL task is a master participant or a slave participant.

3. The method according to claim 1 or 2, characterized in that The first sample is an intersection of a second sample and a third sample corresponding to the first role of the candidate participating entity, and the second sample is a sample supported by the first device.

4. The method according to any one of claims 1 to 3, characterized in that The method further comprises: The first device sends information of a sample used by the target participating entity to perform the first VFL task to the second device.

5. The method according to any one of claims 1 to 4, characterized in that The target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task is further determined according to at least one of the following: The weight of the feature of the sample corresponding to the role supported by the candidate participating entity, the membership weight corresponding to the candidate participating entity, the membership corresponding to the candidate participating entity, or the role supported by the candidate participating entity in the first VFL.

6. A communication method, characterized in that: The method comprises: The second device obtains information of a third sample corresponding to the first role as a candidate participating entity supporting the first VFL task; The second device determines, based on information about a second sample supported by the first device and information about a third sample corresponding to the first role of the candidate participating entity, information about a first sample corresponding to the first role of the candidate participating entity, wherein the first sample is an intersection of the second sample and the third sample, and the first device is a device that initiates the first VFL task; The second device sends information of a first sample corresponding to the first role of the candidate participating entity to the first device.

7. The method according to claim 6, characterized in that The method further comprises: The second device receives information about a target participating entity participating in the first VFL task and / or information about a role of the target participating entity in the first VFL task from the first device.

8. The method according to claim 7, characterized in that The information of the target participating entity participating in the first VFL task and / or the information of the role of the target participating entity in the first VFL task is determined according to the information of the first sample corresponding to the first role of the candidate participating entity.

9. The method according to claim 7 or 8, characterized in that The role of the target participating entity in the first VFL task is a master participant or a slave participant.

10. The method according to any one of claims 6 to 9, characterized in that The second device obtains information of a candidate participating entity supporting the first VFL task as a third sample corresponding to the first role, including: The second device sends a first request message to the candidate participating entity, where the first request message is used to request obtaining information of a sample corresponding to the first role of the candidate participating entity; The second device receives a first response message from the candidate participating entity, where the first response message includes information about the third sample corresponding to the first role of the candidate participating entity.

11. A communication method, characterized in that: The method comprises: The second device obtains information of a third sample corresponding to the first role as a candidate participating entity supporting the first VFL task; The second device determines the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task based on information of the third sample corresponding to the first role of the candidate participating entity.

12. The method according to claim 11, characterized in that The target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task is further determined according to at least one of the following: The weight of the feature of the sample corresponding to the role supported by the candidate participating entity, the membership weight corresponding to the candidate participating entity, the membership corresponding to the candidate participating entity, or the role supported by the candidate participating entity in the first VFL.

13. The method according to claim 11 or 12, characterized in that The method further comprises: The second device receives first information from the first device, where the first information indicates a method for determining a target participating entity of the first VFL task and / or a role of the target participating entity in the first VFL task; The second device determines, based on information of the third sample corresponding to the first role of the candidate participating entity, a target participating entity for the first VFL task and / or a role of the target participating entity in the first VFL task, including: The second device determines the target participating entity of the first VFL task and / or the role of the target participating entity in the first VFL task based on information of the third sample corresponding to the first role of the candidate participating entity and the first information.

14. The method according to claim 13, characterized in that The second device receiving the first information from the first device includes: The second device receives a second request message from the first device, where the second request message includes the first information and is used to request execution of the first VFL task.

15. The method according to claim 14, characterized in that The method further comprises: The first device sends the first information to the second device.

16. The method according to claim 14 or 15, characterized in that The second request message is also used by the first device to request to participate in the first VFL task.

17. The method according to claim 16, characterized in that The second request message includes information indicating that the first device requests to participate in the first VFL task.

18. The method according to claim 16, characterized in that The first information includes the information indicating that the first device requests to participate in the first VFL task.

19. The method according to any one of claims 16 to 18, characterized in that The information of the third sample corresponding to the first role of the candidate participating entity supporting the first VFL task includes information of the third sample corresponding to the first role of the first device.

20. The method according to claim 19, wherein The information for indicating that the first device requests to participate in the first VFL task includes information of a third sample corresponding to the first role of the first device.

21. The method according to any one of claims 11 to 20, characterized in that The method further comprises: The second device sends information about a target participating entity participating in the first VFL task and / or information about a role of the target participating entity in the first VFL task to the first device.

22. The method according to claim 21, characterized in that The method further comprises: The first device receives information about a target participating entity participating in the first VFL task and / or information about a role of the target participating entity in the first VFL task from the second device.

23. The method according to any one of claims 11 to 22, characterized in that The second device obtains information of a candidate participating entity supporting the first VFL task as a third sample corresponding to the first role, including: The second device sends a first request message to the candidate participating entity, where the first request message is used to request obtaining information of a sample corresponding to the first role of the candidate participating entity; The second device receives a first response message from the candidate participating entity, where the first response message includes information about the third sample corresponding to the first role of the candidate participating entity.

24. A communication device, characterized in that: The method comprises at least one module, wherein the at least one module is configured to execute the method according to any one of claims 1 to 23.

25. A communication device, characterized in that: include: processor; The processor is configured to execute a computer program or instruction so that the method according to any one of claims 1 to 23 is implemented.

26. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program or instruction. When the computer program or instruction is executed by the communication device, the method according to any one of claims 1 to 23 is implemented.

27. A computer program product, characterized in that The device comprises a computer program code, and when the computer program code is run on a communication device, the communication device implements the method according to any one of claims 1 to 23.

Images