Communication method and apparatus

Abstract

A communication method and apparatus, which are used for improving the training performance. The method comprises: receiving first capability information from a first management device, wherein the first capability information indicates training capabilities supported by the first management device and / or a network element device; and on the basis of the first capability information, sending a training request to the first management device, wherein the training request is used for requesting the use of a first training technique to execute a first inference function. The method enables a second management device to acquire training techniques supported by a first management device and / or a network element device, so that the second management device can select, on the basis of requirements, an appropriate training technique for a specific inference function and trigger the first management device and / or the network element device to execute training.

Description

A communication method and apparatus

[0001] Cross-reference to Related Applications

[0002] This application claims priority to the Chinese Patent Application No. 202411397663.1, filed on September 30, 2024, and entitled “A communication method and apparatus”, the content of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0003] The present application relates to the field of communication technology, and in particular to a communication method and apparatus. BACKGROUND

[0004] Artificial intelligence / machine learning (AI / ML) technologies and related applications are being increasingly adopted by a wider range of industries. Currently, AI / ML technologies have introduced new training techniques, including federated learning, generative AI learning, reinforcement learning, distributed learning, and other techniques. How to uniformly control the above training techniques has become a problem to be solved. SUMMARY

[0005] The present application provides a communication method and apparatus for improving training performance.

[0006] In a first aspect, the present application provides a communication method. The execution subject of the method can be a second management device or a chip or circuit on the side of the second management device. Taking the second management device as an example, the method comprises: receiving first capability information from a first management device, the first capability information indicating a training capability supported by the first management device and / or a network element device; and sending a training request to the first management device according to the first capability information, the training request being used to request the first management device to perform a first inference function using a first training technique.

[0007] The present application proposes a unified control scheme of training techniques, so that the second management device (for example, a cross-domain management unit) can obtain the training techniques supported by the first management device (for example, a domain management entity) and / or a network element device. Therefore, the second management device (for example, a cross-domain management unit) can select a suitable training technique for a specific inference function according to the demand, trigger the first management device (for example, a domain management entity) and / or a network element device to perform training, which is beneficial to reduce the training overhead of the inference function (or inference service), improve the training speed, accuracy, and the like.

[0008] In a possible design, the training technique includes at least one of the following: federated training, distributed training, reinforcement training, or generative AI.

[0009] In a possible design, the first capability information includes at least one of the following: a name of a training technique supported by the first management device and / or the network element device, a corresponding role of the training technique supported by the first management device and / or the network element device, an entity in which the corresponding role of the training technique supported by the first management device and / or the network element device is located, an inference type to which the training technique supported by the first management device and / or the network element device is applicable, an inference function to which the training technique supported by the first management device and / or the network element device is applicable, or a model capability to which the training technique supported by the first management device and / or the network element device is applicable. The above design can make the training technique selected by the second management device more reasonable by refining the reported training capability, thereby further facilitating reduction of training overhead of the inference function (or inference service) and improvement of training speed, accuracy, and the like.

[0010] In a possible design, the first capability information includes a name of a sub-class of the training technique supported by the first management device and / or the network element device. The above design can make the training technique selected by the second management device more reasonable by refining the reported training capability, thereby further facilitating reduction of training overhead of the inference function (or inference service) and improvement of training speed, accuracy, and the like.

[0011] In a possible design, the first capability information includes at least one of the following: a name of at least one of the following sub-classes of federated training: vertical federated training or horizontal federated training; a name of at least one of the following sub-classes of distributed training: data distributed learning or model distributed learning; a name of at least one of the following sub-classes of generative AI: pre-training, fine-tuning training, incremental pre-training, prompt fine-tuning, retrieval enhancement technology, or generative adversarial network technology.

[0012] In a possible design, the first capability information includes at least one of the following: at least one of the following roles of federated training: a federated training server role or a federated training agent role; at least one of the following roles of reinforcement training: a reinforcement learning agent or a reinforcement learning environment; at least one of the following roles of distributed training: a distributed learning delegator or a distributed learning agent; at least one of the following roles of generative AI: a generator or a discriminator.

[0013] In a possible design, the inference type includes at least one of the following: a management data analytics (MDA) inference function, a radio access network (RAN) intelligence inference function, or a network data analytics function (NWDAF) inference function.

[0014] In a possible design, the inference function includes at least one of the following: at least one inference function of the MDA inference function, including coverage analysis capability or mobility analysis capability; at least one inference function of the RAN intelligent inference function, including random access channel (RACH) optimization, mobile load balancing (MLB) optimization, or energy saving (ES) optimization; at least one inference function of the NWDAF inference function, including slice load analysis or user data congestion analysis.

[0015] In a possible design, the model capability includes at least one of the following: a one-sided model, a two-sided model, a network device model, or a user equipment model.

[0016] In a possible design, the training request includes at least one of the following: a name of the first training technique, a role corresponding to the first training technique, an entity in which the role corresponding to the first training technique is located, or a model capability of the first training technique.

[0017] In a possible design, the method further includes: sending, to the first management device, a capability request, where the capability request is used to request a training capability supported by the first management device and / or the network element device. With the above design, the first management device can report the training capability when needed (that is, when requested by the second management device), which can improve the real-time performance of reporting the training capability, and can reduce the reporting overhead compared with other reporting manners such as periodic reporting.

[0018] In a possible design, the method further includes: selecting the first training technique based on at least one of the following: a computing resource, a number of training data samples, a dimension of training data, a training data sample overlap rate, a data dimension, a training data feature overlap rate, a first inference function, a fault tolerance requirement, or a robustness requirement. With the above design, the rationality of selecting the training technique can be improved.

[0019] In a second aspect, a communication method is provided. An execution subject of the method can be a first management device, or a chip or circuit on the first management device. Taking the first management device as an example, the method includes: sending, to a second management device, first capability information, where the first capability information indicates a training capability supported by the first management device and / or a network element device; and receiving a training request from the second management device, where the training request is used to request that a first inference function is performed by using a first training technique.

[0020] The application proposes a unified control scheme of training technology, so that the second management device (such as a cross-domain management unit) can obtain the training technology supported by the first management device (such as a domain management entity) and / or the network element device, so that the second management device (such as a cross-domain management unit) can select a suitable training technology for a specific inference function according to the demand, trigger the first management device (such as a domain management entity) and / or the network element device to perform training, and reduce the training overhead of the inference function (or inference service), improve the training speed, accuracy, etc.

[0021] In a possible design, the method further includes training the AI / ML model by using the first training technology. The above manner can improve the rationality of the inference function by performing training according to the training technology selected by the second management device.

[0022] In a possible design, the method further includes triggering the network element device to train the AI / ML model by using the first training technology according to the training request. The above manner enables the network element device to perform training according to the training technology selected by the second management device, and can improve the rationality of the inference function.

[0023] In a possible design, the training technology includes at least one of the following: federated training, distributed training, reinforcement training, or generative AI.

[0024] In a possible design, the first capability information includes at least one of the following: the name of the training technology supported by the first management device and / or the network element device, the respective roles of the training technology supported by the first management device and / or the network element device, the respective entities in which the roles of the training technology supported by the first management device and / or the network element device are located, the respective inference types to which the training technology supported by the first management device and / or the network element device is applicable, the respective inference functions to which the training technology supported by the first management device and / or the network element device is applicable, or the respective model capabilities to which the training technology supported by the first management device and / or the network element device is applicable. The above design can make the training technology selected by the second management device more reasonable by refining the reported training capability, thereby further reducing the training overhead of the inference function (or inference service), improving the training speed, accuracy, etc.

[0025] In a possible design, the first capability information includes the name of the sub-class of the training technology supported by the first management device and / or the network element device. The above design can make the training technology selected by the second management device more reasonable by refining the reported training capability, thereby further reducing the training overhead of the inference function (or inference service), improving the training speed, accuracy, etc.

[0026] In a possible design, the first capability information includes at least one of the following: a name of at least one subcategory of federated training: vertical federated training or horizontal federated training; a name of at least one subcategory of distributed training: data distributed learning or model distributed learning; a name of at least one subcategory of generative AI: pre-training, fine-tuning training, incremental pre-training, prompt fine-tuning, retrieval augmentation technology, or generative adversarial network technology.

[0027] In a possible design, the first capability information includes at least one of the following: at least one role of federated training: a federated training server role or a federated training agent role; at least one role of reinforcement training: a reinforcement learning agent or a reinforcement learning environment; at least one role of distributed training: a distributed learning commissioner or a distributed learning agent; at least one role of generative AI: a generator or a discriminator.

[0028] In a possible design, the inference type includes at least one of the following: a management data analytics (MDA) inference function, a radio access network (RAN) intelligence inference function, or a network data analytics function (NWDAF) inference function.

[0029] In a possible design, the inference function includes at least one of the following: at least one inference function of the MDA inference function: a coverage analysis capability or a mobility analysis capability; at least one inference function of the RAN intelligence inference function: a random access channel (RACH) optimization, a mobile load balancing (MLB) optimization, or an energy saving (ES) optimization; at least one inference function of the NWDAF inference function: a slice load analysis or a user data congestion analysis.

[0030] In a possible design, the model capability includes at least one of the following: a one-sided model, a two-sided model, a network device model, or a user equipment model.

[0031] In a possible design, the training request includes at least one of the following: a name of the first training technology, a role corresponding to the first training technology, an entity in which the role corresponding to the first training technology is located, and a model capability of the first training technology.

[0032] In a possible design, the method further includes: receiving a capability request of a second management device, where the capability request is used to request a training capability supported by the first management device and / or the network element device. Through the foregoing design, the first management device can report the training capability when needed (that is, when the second management device requests), which can improve the real-time performance of reporting the training capability, and can reduce the reporting overhead compared with other reporting manners such as periodic reporting.

[0033] In a possible design, the method further includes: receiving second capability information from the at least one network element device, the second capability information of any network element device indicating a training capability supported by the network element device. Through the above design, the first management device can obtain the training capability of the network element device.

[0034] In a possible design, the method further includes: sending a capability request to the at least one network element device, the capability request being used to request a training capability supported by the at least one network element device. Through the above design, the network element device can report the training capability when needed (i.e., when requested by the first management device), which can improve the real-time performance of reporting the training capability, and can reduce the reporting overhead compared with other reporting manners such as periodic reporting.

[0035] In a third aspect, the present application provides a communication method, an execution subject of the method can be a network element device, or a chip or circuit on the network element device side. Taking the network element device as an example, the method includes: obtaining second capability information, the second capability information indicating a training capability supported by the network element device; and sending the second capability information to a first management device.

[0036] The present application can obtain the training technology supported by the network element device through the proposed unified control scheme of the training technology, so that the second management device (for example, a cross-domain management unit) can select a suitable training technology for a specific inference function according to a requirement, trigger the network element device to perform training, and is beneficial to reduce the training overhead of the inference function (or inference service), and improve the training speed and accuracy.

[0037] In a possible design, the method further includes: receiving a capability request from the first management device, the capability request being used to request a training capability supported by the network element device. Through the above design, the network element device can report the training capability when needed (i.e., when requested by the first management device), which can improve the real-time performance of reporting the training capability, and can reduce the reporting overhead compared with other reporting manners such as periodic reporting.

[0038] In a possible design, the method further includes: receiving a training request, the training request being used to request to perform a first inference function by using a first training technology; and training an AI / ML model by using the first training technology. Through the above manner, the inference function can be improved in rationality by performing training according to the training technology selected by the second management device.

[0039] In a possible design, the training technology includes at least one of the following: federated training, distributed training, reinforcement training, or generative AI.

[0040] In a possible design, the first capability information includes names of sub-classes of training techniques supported by the first management device and / or the network element device. The above design can make the training techniques selected by the second management device more reasonable by refining the reported training capabilities, thereby further facilitating reduction of training overhead of the inference function (or inference service), and improving training speed, accuracy, and the like.

[0041] In a possible design, the first capability information includes names of at least one of the following: at least one sub-class of federated training: vertical federated training, or horizontal federated training; at least one sub-class of distributed training: data distributed learning, or model distributed learning; at least one sub-class of generative AI: pre-training, fine-tuning training, incremental pre-training, prompt fine-tuning, retrieval enhancement technology, or generative adversarial network technology.

[0042] In a possible design, the first capability information includes at least one of the following: at least one role of federated training: federated training server role, or federated training agent role; at least one role of reinforcement training: reinforcement learning agent, or reinforcement learning environment; at least one role of distributed training: distributed learning delegator, or distributed learning agent; at least one role of generative AI: generator, or discriminator.

[0043] In a possible design, the inference type includes at least one of the following: management data analytics (MDA) inference function, radio access network (RAN) intelligence inference function, or network data analytics function (NWDAF) inference function.

[0044] In a possible design, the inference function includes at least one of the following: at least one inference function of the MDA inference function: coverage analysis capability, or mobility analysis capability; at least one inference function of the RAN intelligence inference function: random access channel (RACH) optimization, mobile load balancing (MLB) optimization, or energy saving (ES) optimization; at least one inference function of the NWDAF inference function: slice load analysis, or user data congestion analysis.

[0045] In a possible design, the model capability includes at least one of the following: unilateral model, bilateral model, network device model, or user device model.

[0046] In a possible design, the training request includes at least one of the following: a name of the first training technique, a role corresponding to the first training technique, an entity in which the role corresponding to the first training technique is located, and a model capability of the first training technique.

[0047] In a fourth aspect, the present application provides a communication method, the execution subject of the method can be a second management device or a chip or circuit on the second management device side. Taking the second management device as an example, the method comprises: obtaining requirement information of at least one training technology; and sending the requirement information of the at least one training technology to a first management device.

[0048] The present application proposes a unified management and control scheme for training technologies, so that the second management device (for example, a domain management unit) can configure the requirement information of at least one training technology for the first management device (for example, a domain management entity) according to its requirements (for example, business requirements, management requirements, reasoning requirements, etc.), so that the first management device (for example, a domain management entity) and / or a network element device can select a suitable training technology for a specific reasoning function according to the requirement information of the at least one training technology and perform training, which is beneficial to reduce training overhead, improve training speed, improve the accuracy of reasoning functions, etc.

[0049] In a possible design, the at least one training technology comprises at least one of the following: federated training, distributed training, reinforcement training, or generative AI.

[0050] In a possible design, the requirement information comprises at least one of the following: threshold information, a reasoning type, or a reasoning function. The above design can make the training technology selected by the first management device more reasonable by refining the requirement information, thereby further reducing the training overhead of the reasoning function (or reasoning business), improving the training speed, accuracy, etc.

[0051] In a possible design, the threshold information of the distributed training comprises at least one of the following: threshold information of model distributed learning or threshold information of data distributed learning, wherein the threshold information of model distributed learning comprises a computing resource threshold, and the threshold information of data distributed learning comprises at least one of the following: a training data sample number threshold and a training data dimension threshold; the threshold information of the federated training comprises at least one of the following: threshold information of vertical federated training or threshold information of horizontal federated training, wherein the threshold information of vertical federated training comprises at least one of the following: a training data sample overlap rate threshold, a data dimension threshold, or a training data sample number threshold, and the threshold information of horizontal federated training comprises at least one of the following: a data dimension threshold, a training data feature overlap rate threshold, or a training data sample overlap rate threshold; the threshold information of the generative AI comprises at least one of the following: a computing resource threshold or a training data sample number threshold; and the threshold information of the reinforcement training comprises at least one of the following: a fault tolerance requirement or a robustness requirement.

[0052] In a possible design, the reasoning type comprises at least one of the following: an MDA reasoning function, a RAN intelligent reasoning function, or an NWDAF reasoning function.

[0053] In a possible design, the inference function includes at least one of the following: at least one inference function of the MDA inference function, including coverage analysis capability or mobility analysis capability; at least one capability of the RAN intelligent inference function, including RACH optimization, MLB optimization, or ES optimization; at least one capability of the NWDAF inference function, including slice load analysis or user data congestion analysis.

[0054] In a fifth aspect, the present application provides a communication method, and an execution subject of the method can be a first management device or a chip or circuit at the first management device. Taking the first management device as an example, the method includes: receiving demand information of at least one training technique from a second management device; training an AI / ML model based on the demand information of the at least one training technique; and / or triggering a network element device to train the AI / ML model based on the demand information of the at least one training technique.

[0055] The present application proposes a unified control scheme of training techniques, so that the second management device (for example, a domain management unit) can configure demand information of at least one training technique for the first management device (for example, a domain management entity) according to its demand (for example, a business demand, a management demand, an inference demand, etc.), so that the first management device (for example, the domain management entity) and / or the network element device can select a suitable training technique for a specific inference function and perform training according to the demand information of the at least one training technique, which is beneficial to reduce training overhead, improve training speed, improve accuracy of the inference function, etc.

[0056] In a possible design, the at least one training technique includes at least one of the following: federated training, distributed training, reinforcement training, or generative AI.

[0057] In a possible design, the demand information includes at least one of the following: threshold information, an inference type, or an inference function. The above design can make the training technique selected by the first management device more reasonable by refining the demand information, thereby further being beneficial to reduce training overhead of the inference function (or inference business), improve training speed, accuracy, etc.

[0058] In a possible design, the threshold information of the distributed training includes at least one of the following: threshold information of model distributed learning, or threshold information of data distributed learning, where the threshold information of model distributed learning includes a computing resource threshold, and the threshold information of data distributed learning includes at least one of the following: a training data sample quantity threshold and a training data dimension threshold; the threshold information of the federated training includes at least one of the following: threshold information of vertical federated training, or threshold information of horizontal federated training, where the threshold information of vertical federated training includes at least one of the following: a training data sample overlap rate threshold, a data dimension threshold, or a training data sample quantity threshold, and the threshold information of horizontal federated training includes at least one of the following: a data dimension threshold, a training data feature overlap rate threshold, or a training data sample overlap rate threshold; the threshold information of the generative AI includes at least one of the following: a computing resource threshold, or a training data sample quantity threshold; and the threshold information of the reinforcement training includes at least one of the following: a fault tolerance requirement, or a robustness requirement.

[0059] In a possible design, the inference type includes at least one of the following: an MDA inference function, a RAN intelligentization inference function, or an NWDAF inference function.

[0060] In a possible design, the inference function includes at least one of the following: at least one of the following inference functions of the MDA inference function: a coverage analysis capability, or a mobility analysis capability; at least one of the following capabilities of the RAN intelligentization inference function: RACH optimization, MLB optimization, or ES optimization; and at least one of the following capabilities of the NWDAF inference function: slice load analysis, or user data congestion analysis.

[0061] In a possible design, the AI / ML model is trained based on the requirement information, including: selecting a training technique based on at least one of the following: a computing resource, a training data sample quantity, a training data dimension, a training data sample overlap rate, a data dimension, a training data feature overlap rate, a first inference function, a fault tolerance requirement, or a robustness requirement; and training the AI / ML model based on the training technique. The above design can improve the rationality of selecting the training technique.

[0062] In a sixth aspect, a communication method is provided. The execution subject of the method can be a network element device, or a chip or circuit on the network element device side. Taking the network element device as an example, the method includes: receiving requirement information of at least one training technique from a second management device.

[0063] The application provides a unified control scheme for training techniques, so that the second management device (for example, a domain management unit) can configure requirement information of at least one training technique for the network element device according to its requirements (for example, business requirements, management requirements, inference requirements, etc.), which is beneficial to reducing training overhead, improving training speed, and improving the accuracy of inference functions.

[0064] In a possible design, the method further includes training the AI / ML model based on requirement information of the at least one training technique. The above design enables the network element device to select a suitable training technique for a specific inference function according to requirement information of the at least one training technique and perform training, which is beneficial to reducing training overhead, improving training speed, and improving accuracy of the inference function.

[0065] In a possible design, the at least one training technique includes at least one of federated training, distributed training, reinforcement training, or generative AI.

[0066] In a possible design, the requirement information includes at least one of threshold information, an inference type, or an inference function. The above design refines the requirement information, which enables the first management device to select a more reasonable training technique, thereby further reducing training overhead of the inference function (or inference service), improving training speed, accuracy, and the like.

[0067] In a possible design, the threshold information of the distributed training includes at least one of threshold information of model distributed learning or threshold information of data distributed learning, the threshold information of the model distributed learning includes a computing resource threshold, and the threshold information of the data distributed learning includes at least one of a training data sample quantity threshold and a training data dimension threshold; the threshold information of the federated training includes at least one of threshold information of vertical federated training or threshold information of horizontal federated training, the threshold information of the vertical federated training includes at least one of a training data sample overlap rate threshold, a data dimension threshold, or a training data sample quantity threshold, and the threshold information of the horizontal federated training includes at least one of a data dimension threshold, a training data feature overlap rate threshold, or a training data sample overlap rate threshold; the threshold information of the generative AI includes at least one of a computing resource threshold or a training data sample quantity threshold; and the threshold information of the reinforcement training includes at least one of a fault tolerance requirement or a robustness requirement.

[0068] In a possible design, the inference type includes at least one of an MDA inference function, a RAN intelligent inference function, or a NWDAF inference function.

[0069] In a possible design, the inference function includes at least one of at least one inference function of the MDA inference function: coverage analysis capability or mobility analysis capability; at least one capability of the RAN intelligent inference function: RACH optimization, MLB optimization, or ES optimization; or at least one capability of the NWDAF inference function: slice load analysis or user data congestion analysis.

[0070] In a possible design, the AI / ML model is trained based on requirement information, including: selecting a training technique based on at least one of the following: a computing resource, a number of training data samples, a dimension of training data, a training data sample overlap rate, a data dimension, a training data feature overlap rate, a first inference function, a fault tolerance requirement, or a robustness requirement; and training the AI / ML model based on the training technique. The above design can improve the rationality of selecting a training technique.

[0071] In a seventh aspect, the present disclosure provides a communication apparatus, which implements any of the methods provided in the first aspect or the fourth aspect. The communication apparatus can be implemented in hardware, or in a combination of hardware and software.

[0072] In a possible implementation, the communication apparatus includes a processor configured to support the communication apparatus to perform the corresponding functions of the second management device in the above-described methods. The communication apparatus can further include a memory coupled to the processor, which stores program instructions and data necessary for the communication apparatus. Optionally, the communication apparatus further includes an interface circuit, which supports the communication between the communication apparatus and the first management device and other devices.

[0073] In a possible implementation, the communication apparatus includes corresponding functional modules for implementing the steps in the above-described methods. The functions can be implemented in hardware, or in a combination of hardware and software. The hardware or software includes one or more modules corresponding to the above-described functions.

[0074] In a possible implementation, the communication apparatus includes a processing unit and a communication unit, which can perform the corresponding functions in the above-described method examples, as described in the methods provided in the first aspect or the fourth aspect, which will not be repeated here.

[0075] In an eighth aspect, the present disclosure provides a communication apparatus, which implements any of the methods provided in the second aspect or the fifth aspect. The communication apparatus can be implemented in hardware, or in a combination of hardware and software.

[0076] In a possible implementation, the communication apparatus includes a processor configured to support the communication apparatus to perform the corresponding functions of the first management device in the above-described methods. The communication apparatus can further include a memory coupled to the processor, which stores program instructions and data necessary for the communication apparatus. Optionally, the communication apparatus further includes an interface circuit, which supports the communication between the communication apparatus and the second management device, a network element device, and other devices.

[0077] In a possible implementation, the communication apparatus includes respective functional modules for implementing the steps in the above method. The functions can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions.

[0078] In a possible implementation, the communication apparatus includes a processing unit and a communication unit in its structure, which can perform the corresponding functions in the above method examples, see the description of the method in the second aspect or the fifth aspect, which will not be repeated here.

[0079] In a possible implementation, the communication apparatus includes a processing unit and a communication unit in its structure, which can perform the corresponding functions in the above method examples, see the description of the method in the second aspect or the fifth aspect, which will not be repeated here.

[0080] In a possible implementation, the communication apparatus includes a processor configured to support the communication apparatus to perform the corresponding functions of the network element device in the above method. The communication apparatus can also include a memory coupled to the processor, which stores the necessary program instructions and data of the communication apparatus. Optionally, the communication apparatus also includes an interface circuit for supporting the communication between the communication apparatus and other devices such as the first management device.

[0081] In a possible implementation, the communication apparatus includes respective functional modules for implementing the steps in the above method. The functions can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions.

[0082] In a possible implementation, the communication apparatus includes a processing unit and a communication unit in its structure, which can perform the corresponding functions in the above method examples, see the description of the method in the second aspect or the fifth aspect, which will not be repeated here.

[0083] In a possible implementation, the communication apparatus includes a processing unit and a communication unit in its structure, which can perform the corresponding functions in the above method examples, see the description of the method in the second aspect or the fifth aspect, which will not be repeated here.

[0084] In an eleventh aspect, a communication apparatus is provided, which comprises a processor and an interface circuit, the interface circuit being configured to receive signals from other communication apparatuses outside the communication apparatus and transmit the signals to the processor or send signals from the processor to other communication apparatuses outside the communication apparatus, and the processor being configured to implement the method in any possible design of the second aspect or the fifth aspect by means of logic circuit or by executing code instruction.

[0085] In a twelfth aspect, a communication apparatus is provided, which comprises a processor and an interface circuit, the interface circuit being configured to receive signals from other communication apparatuses outside the communication apparatus and transmit the signals to the processor or send signals from the processor to other communication apparatuses outside the communication apparatus, and the processor being configured to implement the method in any possible design of the third aspect or the sixth aspect by means of logic circuit or by executing code instruction.

[0086] In a thirteenth aspect, a computer readable storage medium is provided, which stores a computer program or instruction, and when the computer program or instruction is executed by a processor, the method in any possible design of any one of the first aspect to the sixth aspect is implemented.

[0087] In a fourteenth aspect, a chip system is provided, which comprises a processor and can further comprise a memory, and is configured to implement the method in any possible design of any one of the first aspect to the sixth aspect. The chip system can be composed of a chip or can comprise a chip and other discrete devices.

[0088] In a fifteenth aspect, a communication system is provided, which comprises the apparatus (such as the second management device) in the first aspect, the apparatus (such as the first management device) in the second aspect and the apparatus (such as the network element device) in the third aspect.

[0089] In a sixteenth aspect, a communication system is provided, which comprises the apparatus (such as the second management device) in the fourth aspect, the apparatus (such as the first management device) in the fifth aspect and the apparatus (such as the network element device) in the sixth aspect.

[0090] The technical effects achieved by the technical solutions in any one of the seventh aspect to the sixteenth aspect can be described with reference to the technical effects achieved by the technical solutions in the first aspect, and the repeated parts will not be described herein. BRIEF DESCRIPTION OF DRAWINGS

[0091] FIG. 1 is a schematic diagram of an architecture of a service system provided by the present application;

[0092] FIG. 2 is a schematic diagram of another architecture of a service system provided by the present application;

[0093] FIG. 3 is a flow diagram of a communication method according to an embodiment of the present application;

[0094] FIG. 4 is a flow diagram of another communication method according to an embodiment of the present application;

[0095] FIG. 5 is a structural diagram of a communication apparatus according to an embodiment of the present application;

[0096] FIG. 6 is a structural diagram of another communication apparatus according to an embodiment of the present application. DETAILED DESCRIPTION

[0097] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0098] The embodiments of the present application can be applied to various mobile communication systems, such as a new radio (NR) system, a long term evolution (LTE) system, an advanced long term evolution (LTE-A) system, a future communication system, and other communication systems. Specifically, the embodiments of the present application are not limited thereto. For example, the embodiments of the present application can be applied to a network management architecture of the NR. The network management architecture of the NR can include a management function (MnF). The MnF is a management entity defined by the 3rd generation partnership project (3GPP), and its externally visible behavior and interface are defined as management services (MnS). In a management architecture that gives services, the MnF acts as a MnS producer or a MnS consumer. The MnS produced by the MnS producer of the MnF can have multiple MnS consumers. The MnF can consume multiple management services from one or more management service producers. As shown in FIG. 1, the MnS provided by the MnF can be used to provide services for another MnF (e.g., MnF #A), and in this case, the MnF can act as a MnS producer, and the MnF #A can act as a MnS consumer. In addition, the MnF can also obtain the MnS provided by another MnF (e.g., MnF #B, which can be the same as or different from the MnF #A) from the MnF #B, and in this case, the MnF shown in FIG. 1 can also act as a MnS consumer of the MnS provided by the MnF #B, that is, the same MnF can act as a consumer of the MnS and a producer of the MnS.

[0099] As shown in FIG. 2, it is a schematic diagram of a service-oriented management architecture applicable to the embodiments of the method of the present application. The service-oriented management architecture includes a business support system (BSS), a cross-domain management function unit (CD-MnF), a domain management function unit (Domain-MnF), and a network element (NE). It should be understood that the cross-domain management function unit can be a network management system (NMS), a MnS Producer, a MnS Consumer, or the like, and the domain management function unit can be a wireless automation engine (MAE), an element management system (EMS), a MnS Producer, a MnS Consumer, or the like.

[0100] In the present application, the cross-domain management function unit is used to manage one or more domain management function units. The domain management function unit can be used to manage one or more network elements. The following briefly introduces each unit.

[0101] If the management service is a management service provided by the cross-domain management function unit, the cross-domain management function unit is a management service producer, and the business support system is a management service consumer.

[0102] If the management service is a management service provided by the domain management function unit, the domain management function unit is a management service producer, and the cross-domain management function unit is a management service consumer.

[0103] If the management service is a management service provided by the network element, the network element is a management service producer, and the domain management function unit is a management service consumer.

[0104] The business support system is oriented to communication services, and is used to provide charging, settlement, accounting, customer service, business, network monitoring, communication service life cycle management, business intent translation, and the like. The business support system can be an operator's operation system or a vertical OT system.

[0105] The cross-domain management function unit can also be referred to as a network management function unit (NMF), for example, can be an NMS, a network function management service consumer (NFMS_C), and the like. The cross-domain management function unit provides one or more of the following management functions or management services: network lifecycle management, network deployment, network fault management, network performance management, network configuration management, network assurance, network optimization function, and translation of an intent from a communication service provider (Intent-CSP) of a service producer, and the like.

[0106] The network referred to in the above management functions or management services can include one or more network elements or sub-networks, or can be a network slice. That is, the network management function unit can be a network slice management function unit (NSMF), or a cross-domain management data analysis function unit (MDAF), or a cross-domain self-organization network function (SON Function), or an intent driven management service (Intent Driven MnS).

[0107] In various embodiments of the present application, a network element is an entity that provides network services. The network element can include a core network element, a radio access network element, or a transport network element, and the like. For example, in the architecture shown in FIG. 2, the domain management function unit can include a radio access network domain management function unit, a core network element domain management function unit, or a transport network domain management function unit, wherein the radio access network domain management function unit can be used to manage a radio access network element, the core network element domain management function unit can be used to manage a core network element, and the transport network domain management function unit can be used to manage a transport network element.

[0108] Exemplarily, the core network network element can include, but is not limited to, an access and mobility management function (AMF) entity, a session management function (SMF) entity, a policy control function (PCF) entity, a network data analysis function (NWDAF) entity, a network repository function (NRF), a gateway, and the like.

[0109] The radio access network network element can include, but is not limited to, various types of base stations (for example, a next generation base station (gNB), an evolved Node B (eNB), a central unit control panel (CUCP), a central unit (CU), a distributed unit (DU), a central unit user panel (CUUP), and the like). In the present application, the network function NF is also referred to as the network element NE.

[0110] Optionally, in some deployment scenarios, the cross-domain management function unit can also provide lifecycle management of a subnetwork, deployment of a subnetwork, fault management of a subnetwork, performance management of a subnetwork, configuration management of a subnetwork, assurance of a subnetwork, optimization function of a subnetwork, translation of an intent from a communication service producer or an intent from a communication service consumer (Intent-CSC) of a subnetwork, and the like. The subnetwork here is composed of multiple small subnetworks, which can be a network slice subnetwork.

[0111] The domain management function unit can also be referred to as an NMF or a network element management function unit. For example, the domain management function unit can be an MAE, an EMS, a network function management service producer (NFMS_P), and the like.

[0112] The domain management function unit provides one or more of the following functions or management services: lifecycle management of a subnetwork or a network element, deployment of a subnetwork or a network element, fault management of a subnetwork or a network element, performance management of a subnetwork or a network element, assurance of a subnetwork or a network element, optimization function of a subnetwork or a network element, and translation of an intent from a network operator (Intent-NOP) of a subnetwork or a network element, etc. The subnetwork herein includes one or more network elements. The subnetwork can also include a subnetwork, i.e., one or more subnetworks form a larger subnetwork.

[0113] It should be understood that the intent in this application can be a desire for an intent producer (such as a network element), a system (such as a network or a subnetwork, etc.) in which the intent producer is located, and can include a requirement, a target, or a constraint, etc. The translation of the intent refers to a process of determining a policy for the intent, for example, the policy can be a condition for indicating that the intent is not met. For example, when the intent is energy saving, policy A can be that when the energy consumption is greater than a first threshold, the energy consumption is abnormal (i.e., not energy saving); and policy B can be that when the energy consumption is greater than a second threshold, the energy consumption is abnormal (i.e., not energy saving). It can be understood that even for the same intent, the scheme determined by using different policies to meet the intent can be different.

[0114] Optionally, the subnetwork herein can also be a network slice subnetwork. The domain management system can be a network slice subnetwork management function (NSSMF), a domain management data analysis function (domain MDAF), a domain self-organization network function (SON Function), a domain intent management function, etc.

[0115] The domain management function unit can be classified in the following manner, including:

[0116] According to the network type classification, the domain management function units can be classified into a radio access network (RAN) domain management function (RAN domain MnF), a core network domain management function (CN domain MnF), a transport network domain management function (TN domain MnF), and the like. It should be noted that the domain management function unit can also be a domain network management system, which can manage one or more of the access network, the core network, or the transport network.

[0117] According to the administrative region classification, the domain management function units can be classified into a domain management function unit of a certain region, such as an A city domain management function unit, a B city domain management function unit, and the like.

[0118] A network element is an entity that provides network services, including a core network element and an access network element. The core network element includes an access and mobility management function (AMF), a session management function (SMF), a policy control function (PCF), a network data analytical function (NWDAF), a network repository function (NRF), a gateway, and the like. The access network element includes a base station (such as a gNB or an eNB), a central unit control plane (CUCP), a central unit (CU), a distribution unit (DU), a central unit user plane (CUUP), and the like.

[0119] The network element can provide one or more of the following management functions or services: lifecycle management of the network element, deployment of the network element, fault management of the network element, performance management of the network element, assurance of the network element, optimization function of the network element, and translation of the network element intent, and the like.

[0120] AI / ML techniques and related applications are being increasingly adopted by a wider range of industries, and AI / ML capabilities are being used in various areas of 5GS, including intelligent optimization use cases applied at base stations of RAN, such as mobility load balancing (MLB), mobility robustness optimization (MRO), and energy saving (ES), management data analytics services applied at network management, such as management data analytics (MDA), and network data analytics services applied at core network, such as network data analytics function (NWDAF), etc.

[0121] Currently, the management workflow of AI / ML is defined, including training phase, testing phase, simulation phase, deployment phase, and inference phase.

[0122] In the training phase, one or a group of ML models are trained, including initial training and retraining. It also includes the validation of ML entities to evaluate the performance of ML entities on training data and validation data.

[0123] In the testing phase, if the validation result is not as expected, such as unacceptable variance, the ML model associated with the ML entity needs to be retrained. The training phase is the initial phase of the management workflow of AI / ML.

[0124] In the simulation phase, the ML entity for inference is run in a simulation environment. The purpose is to evaluate the inference performance of the ML entity in the simulation environment before applying it to the target network or system.

[0125] In the deployment phase, the trained ML entity is made available for the process of target AI / ML inference function.

[0126] In the inference phase, the process of using ML entities for inference through AI / ML inference function.

[0127] Currently, the study of R19 version in 3GPP has started, and new training techniques are defined in TR 28.858, such as federated learning, generative AI learning, reinforcement learning, distributed learning, etc. How operators can uniformly control the above training techniques is a problem to be solved.

[0128] In view of this, the embodiments of the present application provide a communication method and device. In the method, the cross-domain management function unit can obtain the training technology supported by the domain management function unit, so as to select the appropriate training technology to trigger the domain management function unit to create a suitable AI / ML model. Alternatively, the cross-domain management function unit sends the requirement of the training technology to the domain management function unit, so that the domain management function unit selects the appropriate training technology to create a suitable AI / ML model. The method and device are based on the same technical concept. Since the principles of the method and device for solving problems are similar, the implementation of the device and the method can be mutually referred to, and the repeated parts will not be described.

[0129] It should be noted that the term "AI / ML" in the present application can be replaced by "AI", "ML", or "AIML", etc. AI / ML can be understood as AI and ML, or AI or ML, or AI and / or ML.

[0130] In the following, a communication method provided by the embodiments of the present application is introduced. In the embodiments of the present application, the method can be applicable to a scenario where both the training function and the inference function are deployed on the domain management function unit, such as MDA, or a scenario where the training function is deployed on the domain management function unit and the inference function is deployed on the base station, such as SON or RAN intelligence. The method can also be applicable to a scenario where both the training function and the inference function are deployed on the base station, such as SON or RAN intelligence. The method can also be applicable to a scenario where the training function is deployed on the cross-domain management function unit and the inference function is deployed on the management function unit or the base station. The function management scenario defined in 3GPP TS28.105 V18.2.0 4a.2 can be referred to, and the present application will not be described in detail. In the above scenarios, the cross-domain management function unit can act as a service consumer of the domain management function unit to manage and control the functions in the domain management function unit.

[0131] In the embodiments of the present application, the first management device can be an AI / ML MnS producer, or a component (such as a chip or a chip system, etc.) in the AI / ML MnS producer. The second management device can be an AI / ML MnS consumer, or a component (such as a chip or a chip system, etc.) in the AI / ML MnS consumer. The network element device can be a network device, a core network element, or a component (such as a chip or a chip system, etc.) in the network element. The first management device and the second management device can be deployed in different entities, or can also be deployed in the same entity, as shown in FIG. 1. In order to facilitate the understanding of the embodiments of the present application, the following examples are taken that the first management device and the second management device are deployed in different entities. It can be understood that in the embodiments of the present application, the role of providing the management service (MnS) of AI / ML capability is called producer, and the role of calling the management service of AI / ML capability is called consumer.

[0132] In an example, the first management device can be a domain management function unit in FIG. 2, the second management device can be a cross-domain management function unit in FIG. 2, and the network element device can be a network element in FIG. 2. The descriptions of the cross-domain management function unit, the domain management function unit, and the network element can refer to the related contents of FIG. 2, which will not be repeated here.

[0133] Referring to FIG. 3, an example flow chart of a communication method provided by the embodiments of the present application is shown, which includes the following steps:

[0134] 301, the first management device sends the first capability information to the second management device. Correspondingly, the first capability information of the first management device is received.

[0135] The first capability information indicates the training capability supported by the first management device and / or the network element device. The network element device can be a network element device managed by the first management device, and the number of the network element device can be one or more, which is not limited here.

[0136] Optionally, the first capability information can include at least one of the following: the name of the training technology supported by the first management device and / or the network element device, the role corresponding to the training technology supported by the first management device and / or the network element device, the entity where the role corresponding to the training technology supported by the first management device and / or the network element device is located, the inference type to which the training technology supported by the first management device and / or the network element device is applicable, the inference function to which the training technology supported by the first management device and / or the network element device is applicable, or the model capability to which the training technology supported by the first management device and / or the network element device is applicable.

[0137] The above-mentioned information will be introduced respectively as follows.

[0138] 1) the name of the training technique supported by the first management device and / or network element device

[0139] For example, the first capability information can include the name of at least one training technique of: federated learning (FL), distributed learning (DL), reinforcement learning (RL), or generative AI.

[0140] Further, the first capability information can include the name of a sub-class of the supported training technique. For example, taking federated learning as an example, the first capability information can include the name of at least one sub-class of federated learning of: vertical federated learning (VFL), or horizontal federated learning (HFL). Taking distributed learning as an example, the first capability information can include the name of at least one sub-class of distributed learning of: data-distributed learning, or model-distributed learning. Taking generative AI as an example, the first capability information can include the name of at least one sub-class of generative AI of: Pre-Training, Fine-Tuning, Incremental Pre-training, prompt-tuning, retrieval augmented generation (RAG), or generative adversarial networks (GANs).

[0141] 2) the respective roles of the training techniques supported by the first management device and / or network element device.

[0142] For example, taking federated learning as an example, the respective roles of the training techniques include a federated learning server role (FL server) and a federated learning client role (FL client). Taking reinforcement learning as an example, the respective roles of the training techniques include a reinforcement learning agent role (RL agent) and a reinforcement learning environment role (RL environment). Taking distributed learning as an example, the respective roles of the training techniques include a distributed learning delegator role (DL delegator) and a distributed learning client role (DL client). Taking the GAN technique of generative AI as an example, the respective roles of the training techniques include a Generator role and a Discriminator role.

[0143] For example, the first capability information can include a name of federated training and a role corresponding to the federated training.

[0144] The first capability information can include a name of reinforcement training and a role corresponding to the reinforcement training.

[0145] The first capability information can include a name of distributed training and a role corresponding to the distributed training.

[0146] The first capability information can include a name of generative AI and a role corresponding to the generative AI.

[0147] 3) an entity in which a role corresponding to a training technology supported by the first management device and / or the network element device is located.

[0148] For example, the entity can be a network element (such as a base station, a core network element, etc.), and can also be a domain management function unit, which is not limited in the present application. The entity can be identified by an identifier of the network element (such as a network device identifier, a cell identifier, a core network element identifier, etc.), or an identifier of the domain management function unit (such as a manufacturer identifier, etc.).

[0149] 4) an inference type to which a training technology supported by the first management device and / or the network element device is respectively applicable

[0150] For example, the inference type includes at least one of an inference function supported by a domain management function unit, an inference function supported by a wireless network element, or an inference function supported by a core network element. The inference function supported by the domain management function unit can be an MDA inference function, the inference function supported by the wireless network element can be a RAN intelligent inference function, and the inference function supported by the core network element can be an NWDAF inference function.

[0151] For example, the first capability information can include a name of federated training and an inference type corresponding to the federated training. The inference type corresponding to the federated training can be understood as an inference type to which the federated training is applicable. For example, the federated training is applicable to the NWDAF inference function.

[0152] The first capability information can include a name of reinforcement training and an inference type corresponding to the reinforcement training. The inference type corresponding to the reinforcement training can be understood as an inference type to which the reinforcement training is applicable. For example, the reinforcement training is applicable to the MDA inference function.

[0153] The first capability information can include a name of distributed training and an inference type corresponding to the distributed training. The inference type corresponding to the distributed training can be understood as an inference type to which the distributed training is applicable. For example, the distributed training is applicable to the RAN intelligent inference function.

[0154] The first capability information can comprise a name of the generative AI and an inference type corresponding to the generative AI. The inference type corresponding to the generative AI can be understood as an inference type to which the generative AI is applicable. The generative AI is applicable to the MDA inference function.

[0155] 5) An inference function to which a training technique supported by the first management device and / or the network element device is respectively applicable.

[0156] By way of example, the inference function comprises at least one of: at least one capability under the MDA inference function, at least one capability under the RAN intelligence inference function, or at least one capability under the NWDAF inference function.

[0157] By way of example, the MDA inference function can comprise at least one of: a coverage analysis capability, or a mobility analysis capability, etc. For details, refer to the types of MDA defined in 3GPP TS 28.104.

[0158] The RAN intelligence inference function can comprise at least one of: a random access channel (RACH) optimization, a mobility load balancing (MLB) optimization, or an energy saving (ES) optimization, etc. For details, refer to the use cases defined in 3GPP TS 28.313, or the use cases defined in TS 28.310.

[0159] The NWDAF inference function can comprise at least one of: a slice load analysis, or a user data congestion analysis, etc. For details, refer to the use cases defined in 3GPP TS 23.288.

[0160] By way of example, the first capability information can comprise a name of the federated training and an inference function corresponding to the federated learning. The inference function corresponding to the federated learning can be understood as an inference function to which the federated learning is applicable.

[0161] The first capability information can comprise a name of the reinforcement training and an inference function corresponding to the reinforcement training. The inference function corresponding to the reinforcement training can be understood as an inference function to which the reinforcement training is applicable.

[0162] The first capability information can comprise a name of the distributed training and an inference function corresponding to the distributed training. The inference function corresponding to the distributed training can be understood as an inference function to which the distributed training is applicable.

[0163] The first capability information can comprise a name of the generative AI and an inference type corresponding to the generative AI. The inference type corresponding to the generative AI can be understood as an inference type to which the generative AI is applicable. The generative AI is applicable to the MDA inference function.

[0164] 6) the model capability of the training technique supported by the first management device and / or the network element device respectively.

[0165] For example, the model capability can include at least one of the following: a single-side model, a double-side model, a network device model, or a user device model. The single-side model refers to training the model on one side, such as training the model only on the network device side or training the model only on the terminal device side. The double-side model refers to training and coordinating the model on both the network device and the terminal device sides. The network device model refers to training the model only on the network device. The user device model refers to training the model only on the terminal device.

[0166] For example, the first capability information can include the name of the federated training and the model capability corresponding to the federated learning.

[0167] The first capability information can include the name of the reinforcement training and the model capability corresponding to the reinforcement training.

[0168] The first capability information can include the name of the distributed training and the model capability corresponding to the distributed training.

[0169] The first capability information can include the name of the generative AI and the model capability corresponding to the generative AI.

[0170] Optionally, the first capability information can further include the name or identifier of the device corresponding to the training capability. For example, the first capability information includes the name of the first management device and the training capability supported by the first management device, the name of the network element device 1 and the training capability supported by the network element device 1, the name of the network element device 2 and the training capability supported by the network element device 2, and so on. The training capability supported by each device can include one or more of the above 6 pieces of information.

[0171] Optionally, the first capability information can further include the computing resources corresponding to the first management device and / or the network element device, such as the use of central processing unit (CPU) for network device training.

[0172] In one possible implementation, the first management device can send the first capability information to the second management device in the following manner: the first management device can create an object for AI / ML model training, and configure the first capability information to the object for AI / ML model training.

[0173] The first management device can send the first capability information to the second management device through a notifyMOICreation operation, and the second management device can receive the first capability information. If the supported training capability of the first management device and / or the network element device is modified or changed, the first management device can also send the updated capability information to the second management device through a notifyMOIAttributeValueChanges operation. Of course, the first management device can also send the capability information through other operations, which is not limited in the present application.

[0174] In a possible implementation, the first capability information can be sent by the first management device at the request of the second management device. For example, before S501, the second management device can send a capability request to the first management device, and the capability request is used to request the training capability supported by the first management device and / or the network element device.

[0175] Optionally, the first management device can obtain the training capability supported by the network element device. For example, the first management device can send a capability request to at least one network element device, and the capability request is used to request the training capability supported by the at least one network element device. The at least one network element device respectively sends second capability information to the first management device, wherein the second capability information of any network element device indicates the training capability supported by the network element device. The information structure of the second capability information is similar to that of the first capability information, which will not be described here.

[0176] As an optional way, the first management device obtains the training capability of the network element device, and can forward it to the second management device, that is, the first capability information includes the second capability information of the at least one network element device.

[0177] As another optional way, after the first management device obtains the training capability of the network element device, the first management device can process the training capability of the network element device, for example, aggregate, etc. That is, the first capability information is determined according to the second capability information of the at least one network element device.

[0178] S302, the second management device sends a training request to the first management device. Correspondingly, the first management device receives the training request from the second management device.

[0179] The training request (TrainingRequest) is used to request to execute the first inference function by using the first training technique.

[0180] Optionally, the training request can include at least one of the following: a name of the first training technique, a role corresponding to the first training technique, an entity in which the role corresponding to the first training technique is located, and a model capability of the first training technique. The name of the first training technique, the role, the entity in which the role is located, and the model capability can be referred to the foregoing description and will not be repeated here.

[0181] In a possible implementation, the second management device can select the first training technique based on at least one of the following: the computing resource, the number of training data samples, the dimension of training data, the overlap rate of training data samples, the overlap rate of training data features, the first inference function, the fault tolerance requirement, or the robustness requirement. It should be noted that the number of training data samples refers to the number of data points for training, and each sample is a data point in the data and has one or more data features. The dimension of training data refers to the number of training data features. It can be understood that each training data sample is composed of one or more features, which can be any numerical, categorical, or other types of data. The overlap rate of training data samples is the probability of repetition of data points in the training data samples. The overlap rate of training data features is the probability of repetition of data features in the training data samples.

[0182] In another possible implementation, when the training function is on the network element, the second management device can select a training technique for model training of the network element device based on the area range to which the inference function is applicable. For example, when two or more network element devices are in the same optimization area, the same training technique can be selected for these network element devices. For example, the second management device can determine the training technique applicable to the first inference function based on the first capability information. It can be understood that after the second management device receives the first capability information supported by the first management device, the second management device selects the training technique applicable to the first inference function based on the capability information supported by the first management device.

[0183] If there are multiple training techniques applicable to the first inference function, the first training technique can be further determined based on the computing resource, the number of training data samples, the dimension of training data, the overlap rate of training data samples, the dimension of data, the overlap rate of training data features, the fault tolerance requirement, or the robustness requirement.

[0184] For example, in the case where the computing resource is greater than or equal to the computing resource threshold, the use of model distributed learning training is preferentially triggered.

[0185] In the case where the number of training data samples is greater than or equal to the number of training data sample threshold, the use of data distributed learning training is preferentially triggered.

[0186] In the case where the overlap rate of training data samples is greater than or equal to the overlap rate of training data sample threshold and the dimension of data is greater than or equal to the dimension of data threshold, the use of vertical federation training can be preferentially triggered.

[0187] In the case that the training data feature overlap rate is greater than or equal to the training data feature overlap rate threshold value and the training data sample overlap rate is less than or equal to the training data sample overlap rate threshold value, the use of horizontal federated training can be preferentially triggered.

[0188] In the case that the fault tolerance requirement and the robustness requirement are met, the use of reinforced training can be preferentially triggered.

[0189] It can be understood that, in the above example, the premise of triggering the corresponding training technology is that the training technology applicable to the first inference function includes the training technology. For example, the premise of triggering model distributed learning training is that the training technology applicable to the first inference function includes model distributed learning training, and the like.

[0190] Optionally, the second management device can send the training request to the first management device in the following manner: the second management device requests the first management device to create an object for ML model training. The first management device creates the object for ML model training and configures the training technology selected by the second management device into the object for ML model training.

[0191] In a possible implementation manner, after receiving the training request, the first management device can train the AI / ML model by using the first training technology.

[0192] In another implementation manner, after receiving the training request, the first management device triggers the network element device to train the AI / ML model by using the first training technology according to the training request. For example, the first management device can forward the training request to the network element device. It can be understood that the network element device meets the training request. The network element device can be indicated by the second management device, for example, the name of the network element device for training the AI / ML model is included in the training request. Alternatively, the network element device can be determined by the first management device according to the training request.

[0193] The above two implementation manners can be implemented independently or in combination, and the present application does not make specific limitations.

[0194] Optionally, the first management device can send a training report to the second management device. In a possible manner, the training report can reuse the content in the MLTrainingReport defined in the existing 3GPP TS 28.105, and the present application does not make limitations here.

[0195] The training report includes the training report of the first management device training the AI / ML model and / or the training report of the network element device training the AI / ML model.

[0196] In a possible implementation manner, the first management device can send the training report to the second management device in the following manner: the first management device can create an object for AI / ML model training, and configure the training report to the object for AI / ML model training.

[0197] The first management device can send the training report to the second management device through a notifyMOICreation operation. If the training report of the first management device training the AI / ML model and / or the training report of the network element device training the AI / ML model is modified or changed, the first management device can also send the updated training report to the second management device through a notifyMOIAttributeValueChanges operation. Of course, the first management device can also send the training report through other operations, which are not limited herein.

[0198] The above describes the manner of selecting a training technology and training an AI / ML model. The following describes an implementation manner of information interaction such as reporting of the first capability information and sending of a training request.

[0199] In an example, the above reporting of the first capability information shown in FIG. 3 can be implemented through an existing object class (information object class, IOC), such as implementing the reporting of the first capability information through an object class (MLTrainingFunction<IOC>) enhanced with an ML training function, including a new non-writable attribute supportedLearningTechnology. It should be noted that the reporting of the first capability information can also be implemented through a new object class, which is not limited herein.

[0200] MLTrainingFunction IOC represents the function of the ML model created by the producer (e.g., the first management device), and the consumer (e.g., the second management device) can obtain the training capability supported by the producer (e.g., the first management device) and / or the network element device, for example, by using supportedLearningTechnology to indicate the training capability supported by the vendor network management (e.g., the first management device) and / or the network element device. The training technology can be represented by a data type. The values of the "support qualifier" include mandatory (M), optional (O), conditionally optional (CO), and conditionally mandatory (CM), the value M indicates that the attribute is a mandatory attribute, the value O indicates that the attribute is an optional attribute, the value CM indicates that the attribute is a mandatory attribute when a certain condition is met, and the value CO indicates that the attribute is an optional attribute when a certain condition is met; the value of "whether readable" includes TRUE (T) and FALSE (F), the value T indicates that the attribute is readable by the second management device, and the value F indicates that the attribute is not readable by the second management device; the value of "whether writable" also includes TRUE and FALSE, the value T indicates that the attribute is writable by the second management device, and the value F indicates that the attribute is not writable by the second management device. The above explanation is also applicable to the tables below, and will not be repeated hereinafter.

[0201] For example, the enhanced MLTrainingFunction IOC can be as shown in Table 1.

[0202] Table 1

[0203] Among them, SupportedlearningTechnology is a data type, used to indicate the first capability information, and its corresponding isWritable is F, that is, the parameter is not writable by the second management device, that is, the parameter is configured by the first management device. Taking the content of the first capability information described above as an example, the attributes contained in SupportedlearningTechnology include the following 5 fields, as shown in Table 2.

[0204] Table 2

[0205] supportedlearningTechnologyName indicates the name of the supported training technology. supportedRole indicates the role corresponding to the training technology respectively. supportedInferenceTypeList indicates the inference type corresponding to the training technology. supportedInferenceNameList indicates the name of the inference function corresponding to the training technology. supportedInferenceNameList indicates the model capability corresponding to the training technology respectively.

[0206] In one example, the sending of the training request shown in FIG. 3 can be implemented by an existing IOC, such as by enhancing the object class (MLTrainingRequest<IOC>) of the ML training request to include a new writable attribute LearningTechnology to implement the sending of the training request described above. It should be noted that a new object class can also be used, which is not limited herein.

[0207] MLTrainingRequest<IOC> represents the operation of the consumer (such as the second management device) triggering the creation of the ML model training object of the producer (such as the first management device). The consumer (such as the second management device) can indicate the training request (for example, by identifying the training request through the attribute LearningTechnology) to the producer (such as the first management device) by using this IOC.

[0208] For example, the enhanced MLTrainingRequest<IOC> can be as shown in Table 3.

[0209] Table 3

[0210] LearningTechnology indicates the first training technology selected by the second management device, and the corresponding isWritable is T, that is, the parameter is writable by the second management device and is configured by the second management device.

[0211] Optionally, LearningTechnology can also be represented by a data type, which can include one or more of the following: the name of the training technology, the role corresponding to the training technology, the entity in which the role corresponding to the training technology is located, and the model capability of the training technology.

[0212] The application provides a unified management scheme of training techniques, so that a second management device (for example, a cross-domain management unit) can obtain training techniques supported by a first management device (for example, a domain management entity), thereby the second management device (for example, the cross-domain management unit) can select a suitable training technique for a specific inference function according to a requirement, trigger the first management device (for example, the domain management entity) and / or a network element device to perform training, and the like, thereby reducing training overhead, improving training speed, improving the accuracy of the inference function, and the like.

[0213] Referring to FIG. 4, an exemplary flowchart of another communication method provided by an embodiment of the application is shown. The method comprises the following steps:

[0214] S401, the second management device sends requirement information of at least one training technique to the first management device. Correspondingly, the first management device receives the requirement information of at least one training technique from the second management device.

[0215] Exemplarily, the at least one training technique comprises at least one of federated training, distributed training, reinforcement training, or generative AI.

[0216] Optionally, the requirement information comprises at least one of the following: a name of the training technique, threshold information, an inference type, or an inference function.

[0217] The requirement information is exemplarily described below in combination with the above training techniques.

[0218] 1) Name of the training technique

[0219] For example, the requirement information corresponding to federated learning comprises a name of federated training. The requirement information corresponding to reinforcement training comprises a name of reinforcement training. The requirement information corresponding to distributed training comprises a name of distributed training. The requirement information corresponding to generative AI comprises a name of generative AI.

[0220] 2) Threshold information

[0221] Exemplarily, taking distributed training as an example, the threshold information of distributed training can comprise at least one of the following: threshold information of model distributed learning, or threshold information of data distributed learning.

[0222] The threshold information of model distributed learning comprises a computing resource threshold. Based on this mode, a possible implementation manner is that the first management device can preferentially trigger the use of model distributed learning training when the computing resource is greater than or equal to the computing resource threshold.

[0223] The threshold information of the data distributed learning includes at least one of the following: a training data sample number threshold or a training data dimension threshold. Based on this manner, a possible implementation manner is that the first management device can preferentially trigger the training using the data distributed learning in a case where the number of training data samples is greater than or equal to the training data sample number threshold.

[0224] Taking federated training as an example, the threshold information of the federated training can include at least one of the following: threshold information of vertical federated training or threshold information of horizontal federated training.

[0225] The threshold information of the vertical federated training includes at least one of the following: a training data sample overlap rate threshold, a data dimension threshold or a training data sample number. Based on this manner, a possible implementation manner is that the first management device can preferentially trigger the vertical federated training in a case where the training data sample overlap rate is greater than or equal to the training data sample overlap rate threshold and the data dimension is greater than or equal to the data dimension threshold.

[0226] The threshold information of the horizontal federated training includes at least one of the following: a data dimension threshold, a training data feature overlap rate threshold or a training data sample overlap rate threshold. Based on this manner, a possible implementation manner is that the first management device can preferentially trigger the horizontal federated training in a case where the training data feature overlap rate is greater than or equal to the training data feature overlap rate threshold and the training data sample overlap rate is less than or equal to the training data sample overlap rate threshold.

[0227] Taking generative AI as an example, the threshold information of the generative AI includes at least one of the following: a computing resource threshold or a training data sample number threshold. Based on this manner, a possible implementation manner is that the first management device can preferentially trigger the generative AI in a case where the computing resource is greater than or equal to the computing resource threshold and the number of training data samples is greater than or equal to the training data sample number threshold.

[0228] Taking reinforcement training as an example, the threshold information of the reinforcement training includes at least one of the following: a fault tolerance requirement or a robustness requirement. Based on this manner, a possible implementation manner is that the first management device can preferentially trigger the reinforcement training in a case where the fault tolerance requirement and the robustness requirement are met.

[0229] 3) Inference type

[0230] Illustratively, the inference type includes at least one of the following: an MDA inference function, a RAN intelligentization inference function or an NWDAF inference function.

[0231] The inference type included in the requirement information of one training technique can be understood as the inference type to which the training technique is applicable. For example, the inference type included in the requirement information of federated learning can be understood as the inference type to which federated learning is applicable. The inference type included in the requirement information of reinforcement training can be understood as the inference type to which reinforcement training is applicable. The inference type included in the requirement information of distributed training can be understood as the inference type to which distributed training is applicable. The inference type included in the requirement information of generative AI can be understood as the inference type to which generative AI is applicable.

[0232] 4) Inference function

[0233] The inference function includes at least one of the following: at least one capability under the MDA inference function, at least one capability under the RAN intelligence inference function, or at least one capability under the NWDAF inference function.

[0234] For example, the MDA inference function can include at least one of the following: coverage analysis capability, or mobility analysis capability, etc. For details, refer to the types of MDA defined in 3GPP TS 28.104.

[0235] The RAN intelligence inference function can include at least one of the following: random access channel (RACH) optimization, mobility load balancing (MLB) optimization, or energy saving (ES) optimization, etc. For details, refer to the use cases defined in 3GPP TS 28.313, or the use cases defined in TS 28.310.

[0236] The NWDAF inference function can include at least one of the following: slice load analysis, or user data congestion analysis, etc. For details, refer to the use cases defined in 3GPP TS 23.288.

[0237] The inference function included in the requirement information of one training technique can be understood as the inference function to which the training technique is applicable. For example, the inference function included in the requirement information of federated learning can be understood as the inference function to which federated learning is applicable. The inference function included in the requirement information of reinforcement training can be understood as the inference function to which reinforcement training is applicable. The inference function included in the requirement information of distributed training can be understood as the inference function to which distributed training is applicable. The inference function included in the requirement information of generative AI can be understood as the inference function to which generative AI is applicable.

[0238] Optionally, the second management device can further send the name or identifier of the first inference function to the first management device.

[0239] In a possible implementation, the second management device can send the requirement information of the at least one training technique to the first management device in the following manner: the second management device can request the first management device to create an object for AI / ML model training. The first management device creates the object for AI / ML model training. The second management device configures the requirement information of the at least one training technique into the object for AI / ML model training.

[0240] Optionally, before S401, the second management device can obtain the requirement information of the at least one training technique. The requirement information of the at least one training technique can be obtained by the second management device from other devices, or determined by the second management device, which is not limited here.

[0241] As an optional solution, after receiving the requirement information of the at least one training technique, the second management device can train the AI / ML model based on the requirement information of the at least one training technique.

[0242] In an implementation, the first management device can select the first training technique based on the requirement information of the at least one training technique and at least one of the following information: computing resource, number of training data samples, training data dimension, training data sample overlap rate, data dimension, training data feature overlap rate, first inference function, fault tolerance requirement, or robustness requirement, and train the AI / ML model by using the first training technique.

[0243] For example, the first management device can determine the training technique suitable for the first inference function based on the requirement information of the at least one training technique. If there are multiple training techniques suitable for the first inference function, the first management device can further determine the first training technique based on the computing resource, the number of training data samples, the training data dimension, the training data sample overlap rate, the data dimension, the training data feature overlap rate, the fault tolerance requirement, or the robustness requirement.

[0244] For example, if the computing resource is greater than or equal to a computing resource threshold, the model distributed learning is triggered preferentially.

[0245] If the number of training data samples is greater than or equal to a number of training data samples threshold, the data distributed learning is triggered preferentially.

[0246] If the training data sample overlap rate is greater than or equal to a training data sample overlap rate threshold, and the data dimension is greater than or equal to a data dimension threshold, the vertical federated training can be triggered preferentially.

[0247] If the training data feature overlap rate is greater than or equal to a training data feature overlap rate threshold, and the training data sample overlap rate is less than or equal to a training data sample overlap rate threshold, the horizontal federated training can be triggered preferentially.

[0248] In the case of meeting the fault tolerance requirement and the robustness requirement, the use of the reinforcement training can be triggered preferentially.

[0249] It can be understood that, in the above example, the premise of triggering the corresponding training technology is that the training technology applicable to the first inference function includes the training technology. For example, the premise of triggering the model distributed learning training is that the training technology applicable to the first inference function includes the model distributed learning training, and so on.

[0250] As another optional solution, after receiving the requirement information of the at least one training technology, the second management device can trigger the network element device to train the AI / ML model based on the requirement information of the at least one training technology.

[0251] In an implementation manner, the first management device can determine the first training technology and the network element device for training the AI / ML model based on the requirement information of the at least one training technology. It can be understood that the network element device supports the first training technology. The first management device can send a training request to the network element device, and the training request is used to request to train the AI / ML model by using the first training technology. It can be understood that the training request can refer to the related description of the training request in the method described in FIG. 3, and will not be repeated here.

[0252] The above two optional manners can be implemented independently or in combination, and the present application does not make specific limitations.

[0253] Optionally, the first management device can send a training report to the second management device. In a possible manner, the training report can reuse the content of the MLTrainingReport defined in the existing 3GPP TS28.105 to perform, and the present application does not make limitations here.

[0254] The training report includes the training report of the first management device training the AI / ML model and / or the training report of the network element device training the AI / ML model.

[0255] In a possible implementation manner, the first management device can send the training report to the second management device by the following manner: the first management device can create an object of the AI / ML model training report, and configure the training result to the object of the AI / ML model training report.

[0256] The first management device can send the training report to the second management device through a notifyMOICreation operation. If the training report of the first management device training the AI / ML model and / or the training report of the network element device training the AI / ML model has modification or change, the first management device can also send the updated training report to the second management device through a notifyMOIAttributeValueChanges operation. Of course, the first management device can also send the training report through other operations, which are not limited herein.

[0257] Optionally, the first management device can also send the first capability information to the second management device, the first capability information indicating the training capability supported by the first management device and / or the network element device. The related description of the first capability information, the way of obtaining the training capability supported by the network element device, the sending and updating manner, and the like can be referred to the related description in the method of FIG. 3, which will not be repeated here.

[0258] The above describes the way of selecting the training technology and training the AI / ML model. The implementation of sending the requirement information is described below.

[0259] In one example, the sending of the requirement information shown in FIG. 4 can be implemented through the existing IOC, such as adding a writable attribute LearningTechnologyRequirement to the object class of the enhanced ML training request (MLTrainingRequest<IOC>) to implement the sending of the above requirement information. It should be noted that the sending of the requirement information can also be implemented through a new object, which is not limited herein.

[0260] MLTrainingRequest<IOC> represents the operation of the consumer (such as the second management device) triggering the creation of the ML model training object of the producer (such as the first management device). The consumer (such as the second management device) can indicate the above requirement information (for example, through the attribute LearningTechnologyRequirement) to the producer (such as the first management device) by using this IOC. For example, the enhanced MLTrainingRequest<IOC> can be as shown in Table 4.

[0261] Table 4

[0262] The LearningTechnologyRequirement indicates the above requirement information. For example, the SupportedlearningTechnology can include the following 5 fields, as shown in Table 5.

[0263] Table 5

[0264] learningTechnologyMetrics indicates threshold information, such as one or more thresholds of a computing resource, a number of training data samples, a dimension of training data, a sample overlap rate of training data, a feature overlap rate of training data, fault tolerance, or robustness, and learningTechnologyValue indicates a value of the computing resource, the number of training data samples, the dimension of training data, the sample overlap rate of training data, the feature overlap rate of training data, the fault tolerance, or the robustness. learningTechnologyRole indicates a role of the training technology. modelCapabilities indicates a model capability of the training technology.

[0265] The application proposes a unified control scheme of a training technology, so that a second management device (for example, a cross-domain management unit) can configure at least one demand information of a training technology for a first management device (for example, a domain management entity) according to its demand (for example, a business demand, a management demand, an inference demand, etc.), so that the first management device (for example, the domain management entity) and / or a network element device can select a suitable training technology for a specific inference function and perform training according to the demand information of the at least one training technology, which is beneficial to reduce training overhead, improve training speed, and improve the accuracy of the inference function.

[0266] Based on the same inventive concept as the method embodiment, the application embodiment provides a communication device. The structure of the communication device can be as shown in FIG. 5, which includes a communication unit 501 and a processing unit 502.

[0267] In an implementation manner, the communication device can be specifically used for implementing the method performed by the second management device in the embodiment of FIG. 3. The device can be the second management device itself, or a chip or chip set or part of a chip in the second management device for executing the related method functions. The processing unit 502 is configured to receive first capability information from a first management device through the communication unit 501, the first capability information indicating a training capability supported by the first management device and / or a network element device. The processing unit 502 is further configured to send a training request to the first management device through the communication unit 501 according to the first capability information, the training request being used for requesting to perform a first inference function by using a first training technology.

[0268] Optionally, the processing unit 502 is further configured to send a capability request to the first management device through the communication unit 501, the capability request being used for requesting the training capability supported by the first management device and / or the network element device.

[0269] Optionally, the processing unit 502 is further configured to select the first training technique based on at least one of the following: a computing resource, a number of training data samples, a dimension of training data, a sample overlap rate of training data, a dimension of data, a feature overlap rate of training data, the first inference function, a fault tolerance requirement, or a robustness requirement.

[0270] In an embodiment, the communication apparatus can be specifically configured to implement the method performed by the first management device in the embodiment of FIG. 3. The apparatus can be the first management device itself, or a chip or chip set or part of a chip in the first management device for performing the functions of the related method. The processing unit 502 is configured to send first capability information to a second management device via the communication unit 501, the first capability information indicating training capabilities supported by the first management device and / or a network element device; and receive a training request from the second management device via the communication unit 501, the training request being used to request execution of a first inference function using a first training technique.

[0271] Optionally, the processing unit 502 is further configured to train an artificial intelligence (AI) / machine learning (ML) model using the first training technique.

[0272] Optionally, the processing unit 502 is further configured to trigger the network element device to train an AI / ML model using the first training technique according to the training request.

[0273] Optionally, the processing unit 502 is further configured to receive a capability request of the second management device via the communication unit 501, the capability request being used to request the training capabilities supported by the first management device and / or the network element device.

[0274] Optionally, the processing unit 502 is further configured to receive second capability information of at least one network element device via the communication unit 501, the second capability information of any of the network element devices indicating training capabilities supported by the network element device.

[0275] Optionally, the processing unit 502 is further configured to send a capability request to the at least one network element device via the communication unit 501, the capability request being used to request the training capabilities supported by the at least one network element device.

[0276] In an embodiment, the communication apparatus can be specifically configured to implement the method performed by the second management device in the embodiment of FIG. 4. The apparatus can be the second management device itself, or a chip or chip set or part of a chip in the second management device for performing the functions of the related method. The processing unit 502 is configured to obtain requirement information of at least one training technique; and the communication unit 501 is configured to send the requirement information of the at least one training technique to a first management device.

[0277] In an implementation, the communication apparatus can be specifically used to implement the method performed by the first management device or the network element device in the embodiment of FIG. 4. The apparatus can be the first management device or the network element device itself, or a chip or chip set or part of the chip in the first management device or the network element device for performing the related method functions. The communication unit 501 is configured to receive the requirement information of at least one training technique from the second management device. The processing unit 502 is configured to train the AI / ML model based on the requirement information of the at least one training technique, and / or trigger the network element device to train the AI / ML model based on the requirement information of the at least one training technique.

[0278] For example, the processing unit 502 is specifically configured to select a training technique based on at least one of the following information: computing resources, number of training data samples, training data dimension, training data sample overlap rate, data dimension, training data feature overlap rate, first inference function, fault tolerance requirement, or robustness requirement; and train the AI / ML model based on the training technique.

[0279] The division of the modules in the embodiments of the present application is illustrative, and is only a logical function division. In actual implementation, another division manner can be used. In addition, each functional module in each embodiment of the present application can be integrated in one processor, or can be physically separated, or two or more modules can be integrated in one module. The integrated module can be implemented in the form of hardware or in the form of a software functional module. It can be understood that the functions or implementation of each module in the embodiments of the present application can be further referred to the related description of the method embodiments.

[0280] In one possible manner, the communication apparatus can be as shown in FIG. 6. The apparatus can be a communication device or a chip in the communication device. The communication device can be the first device in the above embodiments, or the second device in the above embodiments. The apparatus includes a processor 601 and a communication interface 602, and can further include a memory 603. The processing unit 502 can be the processor 601. The communication unit 501 can be the communication interface 602. Optionally, the processor 601 and the memory 603 can be integrated together.

[0281] The processor 601 can be a CPU, or a digital processing unit, etc. The communication interface 602 can be a transceiver, or an interface circuit such as a transceiver circuit, or a transceiver chip, etc. The apparatus further comprises a memory 603 for storing programs executed by the processor 601. The memory 603 can be a non-volatile memory such as a hard disk drive (HDD) or a solid-state drive (SSD), etc., or a volatile memory such as a random-access memory (RAM). The memory 603 can be any other medium capable of carrying or storing desired program codes in the form of instructions or data structures and capable of being accessed by a computer, but is not limited thereto.

[0282] The processor 601 is configured to execute the program codes stored in the memory 603, and specifically configured to execute the actions of the processing unit 502 described above, which will not be repeated herein. The communication interface 602 is specifically configured to execute the actions of the communication unit 501 described above, which will not be repeated herein.

[0283] The specific connection medium between the communication interface 602, the processor 601 and the memory 603 is not limited in the embodiments of the present application. In FIG. 6, the memory 603, the processor 601 and the communication interface 602 are connected through a bus 604, which is represented by a thick line in FIG. 6, and the connection mode between other components is only illustrative and is not limited. The bus can be divided into an address bus, a data bus, a control bus, etc. For convenience of representation, only one thick line is used in FIG. 6, but it does not mean that there is only one bus or only one type of bus.

[0284] The embodiments of the present application further provide a computer readable storage medium for storing computer software instructions required for execution by the processor described above, which contains programs required for execution by the processor described above.

[0285] The embodiments of the present application further provide a communication system comprising a communication apparatus for implementing the function of the first management device in the embodiment of FIG. 3, a communication apparatus for implementing the function of the second management device in the embodiment of FIG. 3, and a communication apparatus for implementing the function of the network element device in the embodiment of FIG. 3.

[0286] The embodiments of the present application further provide a communication system comprising a communication apparatus for implementing the function of the first management device in the embodiment of FIG. 4, a communication apparatus for implementing the function of the second management device in the embodiment of FIG. 4, and a communication apparatus for implementing the function of the network element device in the embodiment of FIG. 4.

[0287] Those skilled in the art will appreciate that embodiments of the present application can be readily used as software, hardware, or a combination of software and hardware. In a software embodiment, various software modules in accordance with embodiments of the present application are stored in a memory such as a computer memory or disk storage for use by, or in connection with, the software on the computer system. The software can provide for programs to be transferred to another computer readable medium (e.g., a removable medium, or a medium conveyed through a computer network) for use in a different system.

[0288] The present application is described in reference to the flow diagrams and / or block diagrams of the methods, apparatus (systems) and computer program products according to this application. It will be understood that each block of the flow diagrams and / or block diagrams, and combinations of blocks in the flow diagrams and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flow diagrams and / or block diagrams block or blocks.

[0289] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the function specified in the flow diagrams and / or block diagrams block or blocks.

[0290] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flow diagrams and / or block diagrams block or blocks.

[0291] Obviously, numerous 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 claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A communication method characterized by comprising: comprising: receiving first capability information from a first management device, the first capability information indicating training capabilities supported by the first management device and / or a network element device; sending a training request to the first management device according to the first capability information, the training request being used to request performing a first inference function using a first training technique.

2. The method of claim 1, wherein, The method further comprises: sending a capability request to the first management device, the capability request being used to request training capabilities supported by the first management device and / or the network element device.

3. The method of claim 1 or 2, wherein, The method further comprises: selecting the first training technique based on the first capability information.

4. A communication method characterized by comprising: comprising: sending first capability information to a second management device, the first capability information indicating training capabilities supported by the first management device and / or a network element device; receiving a training request from the second management device, the training request being used to request performing a first inference function using a first training technique.

5. The method of claim 4, wherein, The method further comprises: training an artificial intelligence (AI) / machine learning (ML) model using the first training technique.

6. The method of claim 4 or 5, wherein, The method further comprises: triggering the network element device to train an AI / ML model using the first training technique according to the training request.

7. The method according to any one of claims 4 to 6, wherein, The method further comprises: receiving a capability request from the second management device, the capability request being used to request training capabilities supported by the first management device and / or the network element device.

8. The method according to any one of claims 4 to 7, wherein, The method further comprises: receiving second capability information from the at least one network element device, the second capability information of any of the network element devices indicating training capabilities supported by the network element device.

9. The method of claim 8, wherein, The method further comprises: sending a capability request to the at least one network element device, the capability request being used to request training capabilities supported by the at least one network element device.

10. The method of any one of claims 1-9, wherein, The first capability information comprises at least one of: a name of a training technique supported by the first management device and / or the network element device; a respective role of a training technique supported by the first management device and / or the network element device; an entity where a respective role of a training technique supported by the first management device and / or the network element device is located; an inference type to which a respective training technique supported by the first management device and / or the network element device is applicable; an inference function to which a respective training technique supported by the first management device and / or the network element device is applicable; or a model capability to which a respective training technique supported by the first management device and / or the network element device is applicable.

11. The method of claim 10, wherein, The training technique comprises at least one of: federated training, distributed training, reinforcement training, or generative AI.

12. The method of claim 11, wherein, The first capability information further comprises at least one of: a name of at least one sub-class of the federated training: vertical federated training, or horizontal federated training; a name of at least one sub-class of the distributed training: data distributed learning, or model distributed learning; a name of at least one sub-class of the generative AI: pre-training, fine-tuning training, incremental pre-training, prompt fine-tuning, retrieval-augmentation technique, or generative adversarial network technique.

13. The method of claim 11 or 12, wherein The role corresponding to the federated training includes a federated training server role or a federated training agent role. The role corresponding to the reinforcement training includes a reinforcement learning agent or a reinforcement learning environment. The role corresponding to the distributed training includes a distributed learning commissioner or a distributed learning agent. The role corresponding to the generative AI includes a generator or a discriminator.

14. The method according to any one of claims 11 to 13, wherein, The inference type includes at least one of an MDA inference function, a RAN intelligent inference function, or a NWDAF inference function.

15. The method according to any one of claims 11 to 14, wherein, The inference function includes at least one of: The MDA inference function includes at least one of a coverage analysis capability or a mobility analysis capability. The RAN intelligent inference function includes at least one of RACH optimization, MLB optimization, or ES optimization. The NWDAF inference function includes at least one of slice load analysis or user data congestion analysis.

16. The method of any one of claims 11-15, wherein, The model capability includes at least one of a one-sided model, a two-sided model, a network device model, or a user equipment model.

17. The method of any one of claims 1-16, wherein, The training request includes at least one of the name of the first training technique, the role corresponding to the first training technique, the entity in which the role corresponding to the first training technique is located, or the model capability of the first training technique.

18. A method of communication, comprising: The method includes: Obtaining requirement information of at least one training technique. Sending the requirement information of the at least one training technique to a first management device.

19. The method of claim 18, wherein, The method further includes: Receiving first capability information indicating training capabilities supported by the first management device and / or a network element device.

20. The method of claim 18 or 19, wherein, The at least one training technique includes at least one of federated training, distributed training, reinforcement training, or generative AI.

21. The method of any one of claims 18-20, wherein, The requirement information includes at least one of threshold information, an inference type, or an inference function.

22. The method of claim 21, wherein, The threshold information of the distributed training includes at least one of model distributed learning threshold information or data distributed learning threshold information, wherein the model distributed learning threshold information includes a computing resource threshold, and the data distributed learning threshold information includes at least one of a training data sample number threshold and a training data dimension threshold. The threshold information of the federated training includes at least one of vertical federated training threshold information or horizontal federated training threshold information, wherein the vertical federated training threshold information includes at least one of a training data sample overlap rate threshold, a data dimension threshold, or a training data sample number threshold, and the horizontal federated training threshold information includes at least one of a data dimension threshold, a training data feature overlap rate threshold, or a training data sample overlap rate threshold. The threshold information of the generative AI includes at least one of a computing resource threshold or a training data sample number threshold. The threshold information of the reinforcement training includes at least one of fault tolerance requirements or robustness requirements.

23. The method of claim 21 or 22, wherein, The inference type comprises at least one of: a management data analytics (MDA) inference function, a radio access network (RAN) intelligence inference function, or a network data analytics function (NWDAF) inference function.

24. The method of any one of claims 21-23, wherein, The inference function comprises at least one of: At least one inference function of the MDA inference function: a coverage analysis capability, or a mobility analysis capability; At least one capability of the RAN intelligence inference function: a random access channel (RACH) optimization, a mobile load balancing (MLB) optimization, or an energy saving (ES) optimization; At least one capability of the NWDAF inference function: a slice load analysis, or a user data congestion analysis.

25. A method of communication, comprising: Comprise: Receiving requirement information of at least one training technique from a second management device; Training an artificial intelligence / machine learning (AI / ML) model based on the requirement information of the at least one training technique; And / or, triggering a network element device to train an AI / ML model based on the requirement information of the at least one training technique.

26. The method of claim 25, wherein, The training of the AI / ML model based on the requirement information of the at least one training technique comprises: Selecting a first training technique based on the requirement information of the at least one training technique; Training the AI / ML model based on the first training technique.

27. The method of claim 25, wherein, The triggering of the network element device to train the AI / ML model based on the requirement information of the at least one training technique comprises: Determining a first training technique and the network element device for training the AI / ML model based on the requirement information of the at least one training technique, the network element device supporting the first training technique; Sending a training request to the network element device, the training request being used to request training of the AI / ML model using the first training technique.

28. The method of any one of claims 25-27, wherein, The at least one training technique comprises at least one of: federated training, distributed training, reinforcement training, or generative AI.

29. The method of any one of claims 25-28, wherein, The requirement information comprises at least one of: threshold information, an inference type, or an inference function.

30. The method of claim 29, wherein, The threshold information of the distributed training comprises at least one of: threshold information of model distributed learning, or threshold information of data distributed learning, wherein the threshold information of model distributed learning comprises a computing resource threshold, and the threshold information of data distributed learning comprises at least one of: a training data sample number threshold and a training data dimension threshold; The threshold information of the federated training comprises at least one of: threshold information of vertical federated training, or threshold information of horizontal federated training, wherein the threshold information of vertical federated training comprises at least one of: a training data sample overlap rate threshold, a data dimension threshold, or a training data sample number threshold, and the threshold information of horizontal federated training comprises at least one of: a data dimension threshold, a training data feature overlap rate threshold, or a training data sample overlap rate threshold; The threshold information of the generative AI comprises at least one of: a computing resource threshold, or a training data sample number threshold; The threshold information of the reinforcement training comprises at least one of: a fault tolerance requirement, or a robustness requirement.

31. The method of claim 29 or 30, wherein, The inference type comprises at least one of: a management data analytics (MDA) inference function, a radio access network (RAN) intelligence inference function, or a network data analytics function (NWDAF) inference function.

32. The method of any one of claims 29-31, wherein, The inference function comprises at least one of: at least one inference function of the MDA inference function: a coverage analysis capability, or a mobility analysis capability; at least one capability of the RAN intelligence inference function: a random access channel (RACH) optimization, a mobile load balancing (MLB) optimization, or an energy saving (ES) optimization; at least one capability of the NWDAF inference function: a slice load analysis, or a user data congestion analysis.

33. The method of any one of claims 25-32, wherein, The method further comprises: sending, to the second management device, first capability information indicating a training capability supported by the first management device and / or the network element device.

34. A communications device, characterized by A processor and a memory, the memory being configured to store program instructions, the processor being configured to execute the program instructions to cause the method of any one of claims 1-33 to be performed.

35. A computer readable storage medium, characterized in that, The computer readable storage medium has stored therein computer readable instructions which, when executed on a communication device, cause the method of any one of claims 1-3, 10-17, or the method of any one of claims 4-17, or the method of any one of claims 18-24, or the method of any one of claims 25-33 to be performed.

36. A computer program product, characterised in that, The computer program product, when executed on a device, causes the device to perform the method of any one of claims 1-3, 10-17, or the method of any one of claims 4-17, or the method of any one of claims 18-24, or the method of any one of claims 25-33.

37. A communication system, characterized by The communication system comprises means for performing the method of any one of claims 1-3, 10-17, and means for performing the method of any one of claims 4-17.

38. A communication system, characterized by The communication system comprises means for performing the method of any one of claims 18-24, and means for performing the method of any one of claims 25-33.

Images