Communication method and apparatus

Abstract

Embodiments of the present application provide a communication method and apparatus. The method relates to the technical field of communications. In the method, each network element can be registered in a second network element, so that a certain distributed learning initiator (such as a first network element) can acquire some network elements (such as at least one third network element) meeting distributed learning requirements by means of the second network element, so that the first network element can determine a participant finally participating in the distributed learning from among the obtained at least one third network element. Online determination of distributed participants is realized, and the flexibility of determining the participants is improved.

Description

Communication method and device

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to the Chinese patent application filed with the State Intellectual Property Office of the People's Republic of China on February 16, 2024, with application number 202410180013.5 and application name "A Communication Method and Device", the entire contents of which are incorporated by reference into this application. Technical Field

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

[0004] Distributed learning refers to a learning method that combines the features of multiple participants to jointly train a distributed learning model. The model obtained through distributed learning can be used to handle certain tasks. The initiator of distributed learning receives the task to be processed and determines the features required for the distributed learning model to handle it. Since the initiator may not be able to provide all the features required to train the distributed learning model, it may need to collaborate with other participants in the distributed learning process.

[0005] Currently, the initiator and other participants agree on each participant in advance, that is, each participant is determined offline, and the flexibility of the method of determining the participants is poor. Summary of the Invention

[0006] The present application provides a communication method and apparatus for improving the flexibility of determining participating parties.

[0007] In a first aspect, an embodiment of the present application provides a communication method. The communication method can be executed by a first network element, which can be any type of network element. The first network element can be an existing network element or a newly added network element, etc., without limitation. The first network element can be a device (such as a terminal device or a network device), a software module (such as an application) or a hardware module (such as a chip) in the device, or a client, etc., without specific limitation. The method includes: sending a first request to a second network element, the first request is used to request the discovery of a network element, and the first request indicates one or more features, and the features corresponding to the data used for distributed learning model training of the network element requested to be discovered include one or more features; receiving first information from the second network element, the first information indicating at least one third network element; receiving first information from the second network element, the first information indicating at least one third network element, wherein M third network elements participate in the distributed learning model training, the M third network elements are part or all of the at least one third network element, and M is a positive integer.

[0008] A distributed learning model may also be referred to as a model for short, which refers to a model that is learned (or trained) in a distributed learning manner. A distributed learning model may be a function (such as an artificial intelligence (AI) function or a machine learning (ML) function), a feature, or an algorithm that can be described as distributed learning. The first network element may be regarded as the initiator of the distributed learning model training and as one of the participants. At least one third network element discovered by the first network element through the second network element may be regarded as a candidate participant in the distributed learning model training. The M third network elements may be participants participating in the distributed learning model training, that is, other participants other than the first network element. Optionally, the first network element may determine, based on the first information, the M network elements for the distributed learning model training from the at least one third network element, or directly use the at least one third network element as a network element participating in the distributed learning model training, without limitation.

[0009] In an embodiment of the present application, the first network element can discover at least one third network element through the second network element, and select some or all of the at least one third network element to participate in the distributed learning model training. This provides a way for the first network element to determine the participants without having to pre-negotiate each participant, and provides a way to align the distributed learning model training online. The first network element can determine other participants at any time, that is, on the one hand, the first network element can flexibly determine other participants, and on the other hand, the time for the first network element to determine other participants is also relatively flexible, thereby improving the flexibility of determining the participants. Moreover, due to the higher flexibility in determining the participants, this method can be applied to more scenarios involving determining the participants.

[0010] In one possible implementation, the first request also indicates one or more network element types, the characteristics corresponding to each of the one or more network element types, and the characteristics corresponding to each network element type belong to one or more characteristics; wherein the network element type of at least one third network element belongs to one or more network element types.

[0011] The first request can be understood as (or alternatively described as) used to indicate one or more network element types, and the features corresponding to each network element type in the one or more network element types. The features corresponding to each network element type belong to one or more features, which can be described as: each feature in the one or more features is a feature corresponding to one network element type in the one or more network element types. Among them, the features corresponding to each network element type in the one or more network element types belong to one or more features, that is, the one or more features are the union of the features corresponding to one or more network element types. The features corresponding to one of the network element types can be understood as the features of the network element type specified by the first network element for participating in the training of the distributed learning model.

[0012] In the above implementation, the first request can also specify one or more network element types. In this way, at least one third network element discovered by the first network element also belongs to these one or more network element types. In this way, the M third network elements finally selected are more in line with the needs of the first network element, ensuring the smooth execution of subsequent distributed learning model training.

[0013] In a possible implementation manner, the first information further indicates a feature corresponding to each third network element in the at least one third network element.

[0014] The features corresponding to each third network element may be features supported by each third network element (supported features), features supported by each third network element and belonging to one or more features, or features determined by the second network element for each third network element to participate in the training of the distributed learning model. In the case where the features corresponding to each third network element are features determined by the second network element for each third network element to participate in the training of the distributed learning model, it does not mean that the features corresponding to each third network element will definitely participate in the training of the distributed learning model.

[0015] In the above embodiment, the first network element can not only discover at least one third network element through the first information, but also discover the characteristics corresponding to at least one third network element, so as to facilitate determining the M third network elements ultimately used to participate in the distributed learning model training, and the characteristics of these M third network elements used to participate in the distributed learning model training.

[0016] In a possible implementation manner, the feature corresponding to each third network element in the at least one third network element is a feature among the one or more features.

[0017] Among them, the features corresponding to any two third network elements in at least one third network element may have the same features (that is, there is overlap), or may not have the same features (that is, there is no overlap), and there is no specific limitation on this.

[0018] In the above implementation, the first information does not need to indicate some unnecessary features of the third network element, that is, features that will not be used at all to participate in the distributed learning model training. In this way, the resource overhead of transmitting the first information can be relatively reduced.

[0019] In one possible implementation, the method further includes: determining the characteristics of each of the M third network elements for participating in the distributed learning model training; and sending second information to each of the M third network elements, wherein the second information sent to one of the at least one third network element is used to indicate the characteristics of a third network element for participating in the distributed learning model training.

[0020] In the above implementation, the first network element can not only determine the M network elements used to participate in the distributed learning model training, but also determine the features of the M third network elements used to participate in the distributed learning model training. This aligns the features of each participant in the distributed learning model training and provides a mechanism for feature alignment. Feature alignment can further improve the accuracy of distributed learning model training.

[0021] In one possible implementation, the first information includes information about at least one network element group, wherein the information about one network element group includes characteristics corresponding to the one network element group, wherein the one network element group includes some or all of the network elements in the at least one third network element, and the characteristics corresponding to the one network element group include one or more characteristics. Optionally, the information about one network element group also includes identifiers of the network elements included in the one network element group and / or an identifier of the one network element group.

[0022] In the above embodiment, the second network element can group at least one third network element, and the first information indicates at least one third network element through the information of at least one network element group, and can also indicate the corresponding characteristics of at least one third network element. Since a network element group can be used as a candidate other participant for participating in the distributed learning model training, the first network element can directly select a network element group from at least one network element group as the other participant that ultimately participates in the distributed learning model training. In this way, it is convenient for the first network element to intuitively determine M third network elements, thereby relatively reducing the processing amount of the first network element.

[0023] In one possible implementation, the first request also indicates one or more network element types, and characteristics corresponding to each network element type in the one or more network element types, and the characteristics corresponding to each network element type belong to one or more characteristics; wherein the network element type of a network element group belongs to one network element type in the one or more network element types.

[0024] In the above implementation, by limiting the network element type corresponding to the feature, the provided feature is made more targeted and accurate. For example, if one of the features is the location of user equipment (UE), and it is assumed that both the application function (AF) and the network data analytics function (NWDAF) can obtain UE location information, but the UE location obtained by the AF is more accurate, then this method can be used to specify the AF to provide the UE location.

[0025] In one possible implementation, the method further includes: sending third information to at least one third network element respectively, wherein the third information sent to one of the at least one third network element indicates that one third network element feeds back a characteristic corresponding to the third network element; and receiving fourth information from at least one third network element respectively, wherein the fourth information received from one of the at least one third network element indicates a characteristic corresponding to each third network element.

[0026] In the above embodiment, the first network element can request its corresponding characteristics from at least one third network element on its own, so as to facilitate more accurate acquisition of the characteristics corresponding to at least one third network element, and there is no need for the second network element to notify the first network element of the characteristics corresponding to the at least one third network element, thereby reducing the processing volume of the second network element.

[0027] In a possible implementation manner, the method further includes: sending fifth information to the second network element, where the fifth information indicates features supported by the first network element.

[0028] In the above implementation, the first network element can be registered in the second network element so that the first network element can be discovered by the second network element during training of other distributed learning models.

[0029] In a second aspect, an embodiment of the present application provides a communication method. The communication method can be performed by a second network element. The method can be performed by the second network element, and the second network element can be any type of network element. The second network element can be an existing network element or a newly added network element, etc., without limitation. The second network element can be a device (such as a terminal device or a network device), a software module (such as an application) or a hardware module (such as a chip) in the device, or a client, etc., without specific limitation. The method includes: receiving a first request from a first network element, the first request is used to request discovery of a network element, the first request indicates one or more features, and the features corresponding to the data of the network element requested to be discovered for distributed learning model training include one or more features; sending first information to the first network element, the first information indicating at least one third network element, wherein M third network elements participate in the distributed learning model training, the M third network elements are part or all of the at least one third network element, and M is a positive integer.

[0030] In one possible implementation, the first request also indicates one or more network element types, wherein the characteristics corresponding to each network element type in the one or more network element types belong to one or more characteristics; wherein the network element type of at least one third network element belongs to one or more network element types.

[0031] In a possible implementation manner, the first information further indicates a feature corresponding to each third network element in the at least one third network element.

[0032] In a possible implementation manner, the feature corresponding to each third network element in the at least one third network element is a feature among the one or more features.

[0033] In one possible implementation, the first information includes: information of at least one network element group, wherein the information of one network element group includes characteristics corresponding to one network element group, wherein one network element group includes part or all of at least one third network element, and the characteristics corresponding to one network element group include one or more characteristics.

[0034] In one possible implementation, the first request also indicates one or more network element types, wherein the characteristics corresponding to each network element type in the one or more network element types belong to one or more characteristics; wherein the network element type of a network element group belongs to one network element type in the one or more network element types.

[0035] In a possible implementation manner, the method further includes: receiving fifth information from the first network element, where the fifth information indicates features supported by the first network element.

[0036] In a possible implementation manner, the method further includes: receiving sixth information from at least one third network element, and receiving sixth information from one of the at least one third network element to indicate a feature supported by the third network element.

[0037] In a third aspect, embodiments of the present application provide a communication method. The communication method can be performed by a first network element. The method can be performed by the first network element, and the first network element can be any type of network element, including an existing network element or a newly added network element, without limitation. The first network element can be a device (such as a terminal device or a network device), a software module (such as an application) or a hardware module (such as a chip) in the device, or a client, without specific limitation. The method includes: determining at least one third network element; sending first information to each of the at least one third network elements, wherein the first information sent to one of the at least one third network elements is used to request features supported by the third network element, wherein some or all of the features supported by the at least one third network element are features corresponding to data trained by a distributed learning model; receiving second information from each of the at least one third network element, wherein the second information received from the one of the at least one third network element indicates features supported by the third network element; and determining features for M third network elements to participate in distributed learning model training, wherein the M third network elements are some or all of the at least one third network element, and M is a positive integer.

[0038] In this embodiment of the present application, the first network element can directly determine M third network elements participating in the distributed learning model training from at least one third network element, thereby increasing the flexibility of determining the participating parties. Furthermore, while the at least one third network element is registered with the second network element, the features supported by the at least one third network element are not registered. This reduces the risk of leakage of the features supported by the at least one third network element and ensures the security of the features supported by the at least one third network element.

[0039] In one possible implementation, the method further includes: sending third information to M third network elements respectively, wherein the third information sent to one third network element among at least one third network element indicates a feature of a third network element for participating in distributed learning model training.

[0040] In the above implementation, the first network element may also notify M third network elements of features for participating in distributed learning model training, providing a mechanism for aligning features of distributed learning model training.

[0041] In one possible implementation, determining at least one third network element includes: receiving fourth information from a second network element, where the fourth information indicates information of the at least one third network element; and determining the at least one third network element based on the fourth information.

[0042] In the above implementation manner, the first network element can discover at least one third network element through the second network element, so that the first network element can intuitively and simply discover the at least one third network element.

[0043] In a fourth aspect, an embodiment of the present application provides a communication method. The communication method can be executed by a third network element. The method can be executed by a third network element, and the third network element can be any type of network element. The third network element may be an existing network element or a newly added network element, etc., and there is no limitation on this. The third network element can be a device (such as a terminal device or a network device, etc.), a software module (such as an application) or a hardware module (such as a chip) or a client in the device, etc., and there is no specific limitation on this. The method includes: receiving first information from a first network element, the first information is used to request features supported by the third network element; sending second information to the first network element, the second information indicating features supported by the third network element.

[0044] In a possible implementation, the method further includes: receiving third information from the first network element, where the third information indicates features of the third network element for participating in distributed learning model training.

[0045] In a possible implementation manner, the method further includes: sending fifth information to the second network element, where the fifth information indicates information of the third network element.

[0046] In a fifth aspect, an embodiment of the present application provides a communication device. The communication device may be the first network element in the first aspect above, or a software module or hardware module (e.g., a chip) configured in the first network element. The communication device includes corresponding means (means) or modules for executing the first aspect above or any possible implementation method. For example, the communication device includes a processing module (sometimes also referred to as a processing unit), and a transceiver module (sometimes also referred to as a transceiver unit).

[0047] For example, a transceiver module is used to send a first request to a second network element and receive first information from the second network element, where the first information indicates at least one third network element; wherein M third network elements in the at least one third network element are used to participate in distributed learning model training, and the M third network elements are part or all of the network elements in the at least one third network element.

[0048] In a possible implementation, the communication device may also execute any possible implementation in the first aspect above, which will not be listed one by one here.

[0049] In a sixth aspect, an embodiment of the present application provides a communication device. The communication device may be the second network element in the second aspect above, or a software module or hardware module (e.g., a chip) configured in the second network element. The communication device includes corresponding means (means) or modules for executing the second aspect above or any possible implementation method. For example, the communication device includes a processing module (sometimes also referred to as a processing unit), and a transceiver module (sometimes also referred to as a transceiver unit).

[0050] For example, a transceiver module is used to receive a first request from a first network element, and to send first information to the first network element, where the first information indicates at least one third network element, wherein M third network elements participate in distributed learning model training, and the M third network elements are part or all of the at least one third network element.

[0051] In a possible implementation, the communication device may also execute any possible implementation in the second aspect above, which will not be listed one by one here.

[0052] In a seventh aspect, an embodiment of the present application provides a communication device. The communication device may be the first network element in the third aspect above, or a software module or hardware module (e.g., a chip) configured in the first network element. The communication device includes corresponding means (means) or modules for executing the third aspect above or any possible implementation method. For example, the communication device includes a processing module (sometimes also referred to as a processing unit), and a transceiver module (sometimes also referred to as a transceiver unit).

[0053] For example, a processing module is used to determine at least one third network element; a transceiver module is used to send first information to at least one third network element and receive second information from at least one third network element; the processing module is also used to determine the characteristics of M third network elements for participating in the training of the distributed learning model, and the M third network elements are part or all of the at least one third network element.

[0054] In a possible implementation, the communication device may also execute any possible implementation in the third aspect above, which will not be listed one by one here.

[0055] In an eighth aspect, an embodiment of the present application provides a communication device. The communication device may be the third network element in the fourth aspect, or a software module or hardware module (e.g., a chip) configured in the third network element. The communication device includes corresponding means or modules for executing the fourth aspect or any possible implementation method. For example, the communication device includes a processing module (sometimes also referred to as a processing unit), and a transceiver module (sometimes also referred to as a transceiver unit).

[0056] For example, the transceiver module is configured to receive first information from a first network element, the first information being used to request features supported by a third network element, and to send second information to the first network element, the second information indicating features supported by the third network element.

[0057] In a possible implementation, the communication device can also execute any possible implementation in the fourth aspect above, which will not be listed one by one here.

[0058] In a ninth aspect, an embodiment of the present application provides a communication device, comprising: a processor and a memory, wherein the memory is used to store program instructions, and the processor is used to execute the program instructions in the memory, to perform the method according to any one of the first to fourth aspects.

[0059] Optionally, the wireless communication device further includes a communication interface, and the processor may be coupled to the communication interface, or the processor may be independently provided with the communication interface.

[0060] In one implementation, when the communication apparatus is a wireless communication device, the communication interface may be a transceiver or an input / output interface. Alternatively, the transceiver may be a transceiver circuit. Alternatively, the input / output interface may be an input / output circuit.

[0061] In another implementation, when the communication device is a chip or a chip system, the communication interface may be an input / output interface, an interface circuit, an output circuit, an input circuit, a pin, or related circuits on the chip or chip system. The processor may also be embodied as a processing circuit or a logic circuit.

[0062] Optionally, the communication device further includes other components, such as an antenna, an input / output module or an interface, etc. These components may be hardware, software, or a combination of software and hardware.

[0063] In a tenth aspect, an embodiment of the present application provides a communication device. The communication device includes: a processing circuit and an interface circuit; wherein: the interface circuit is configured to couple with a memory external to the communication device and provide a communication interface for the processing circuit to access the memory; and the processing circuit is configured to execute program instructions in the memory to implement the method according to any one of aspects 1 to 4.

[0064] In a specific implementation, the communication device may be a chip, the input circuit may be an input pin, the output circuit may be an output pin, and the processing circuit may be a transistor, a gate circuit, a trigger, or various logic circuits. The input signal received by the input circuit may be, for example, but not limited to, received and input by a receiver, and the signal output by the output circuit may be, for example, but not limited to, output to and transmitted by a transmitter. The input circuit and the output circuit may be the same circuit, which functions as an input circuit and an output circuit at different times. The embodiments of the present application do not limit the specific implementation of the processor and various circuits.

[0065] In one implementation, the communication device may be a wireless communication device, that is, a computer device that supports wireless communication functions. Specifically, the wireless communication device may be a terminal such as a smartphone, or a wireless access network device such as a base station. The system chip may also be referred to as a system on chip (SoC), or simply as an SoC chip. The communication chip may include a baseband processing chip and a radio frequency processing chip. The baseband processing chip is sometimes also referred to as a modem or baseband chip. The radio frequency processing chip is sometimes also referred to as a radio frequency transceiver or radio frequency chip. In a physical implementation, some or all of the chips in the communication chip may be integrated inside the SoC chip. For example, the baseband processing chip is integrated into the SoC chip, and the radio frequency processing chip is not integrated with the SoC chip. The interface circuit may be the radio frequency processing chip in the wireless communication device, and the processing circuit may be the baseband processing chip in the wireless communication device.

[0066] In another implementation, the communication device may be a component of a wireless communication device, such as an integrated circuit product such as a system-on-chip (SoC) or a communication chip. The interface circuit may be an input / output interface, interface circuit, output circuit, input circuit, pin, or related circuit on the chip or chip system. The processor may also be embodied as a processing circuit or a logic circuit.

[0067] In an eleventh aspect, an embodiment of the present application provides a chip system. The chip system includes a processor and an interface. The processor is configured to call and execute instructions from the interface. When the processor executes the instructions, the method described in any one of aspects 1 to 4 is implemented.

[0068] In a twelfth aspect, an embodiment of the present application provides a computer-readable storage medium for storing a computer program or instruction, which, when executed, implements the method described in any one of the first to fourth aspects above.

[0069] In a thirteenth aspect, an embodiment of the present application provides a computer program product comprising instructions, which, when executed on a computer, implements the method described in any one of the first to fourth aspects above.

[0070] Regarding the beneficial effects of the second, fourth to thirteenth aspects, reference may be made to the beneficial effects discussed in the first or third aspect, which will not be listed here. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] FIG1 is a schematic diagram of data distribution for vertical federated learning applicable to an embodiment of the present application;

[0072] FIG2 is a schematic diagram of a distributed learning scenario applicable to an embodiment of the present application;

[0073] Figure 3 is a schematic diagram of a vertical federated learning process;

[0074] FIG4 is a schematic diagram of a distributed learning scenario provided by an embodiment of the present application;

[0075] FIG5 is a schematic diagram of an intelligent network architecture based on NWDAF provided in an embodiment of the present application;

[0076] FIG6 is a schematic diagram of another distributed learning scenario provided by an embodiment of the present application;

[0077] FIG7 is a schematic diagram of a communication method provided in an embodiment of the present application;

[0078] FIG8 is a schematic diagram of another communication method provided in an embodiment of the present application;

[0079] FIG9 is a schematic diagram of another communication method provided in an embodiment of the present application;

[0080] FIG10 is a schematic diagram of another communication method provided in an embodiment of the present application;

[0081] FIG11 is a schematic diagram of another communication method provided in an embodiment of the present application;

[0082] FIG12 is a schematic structural diagram of a communication device provided in an embodiment of the present application;

[0083] FIG13 is a schematic structural diagram of another communication device provided in an embodiment of the present application;

[0084] FIG14 is a schematic structural diagram of another communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0085] The specific implementation of the present application is described below with reference to the drawings in the embodiments of the present application.

[0086] Below, some terms used in the embodiments of the present application are explained to facilitate understanding by those skilled in the art.

[0087] 1. Data (or dataset)

[0088] The data can be used for model training or inference. The data can be in various forms, such as structured data, such as tables or databases, or unstructured data, such as text, images, audio, or video. For example, the data includes at least one of the service experience, location, average throughput, average packet delay, radio access type (RAT), reference signal receiving power (RSRP), reference signal receiving quality (RSRQ), or radio access network throughput (RAN throughput) of an object (such as a user).

[0089] The data used for model training can be divided into two dimensions (or spaces): the sample space and the feature space. The sample space represents the sample objects, that is, which objects (such as users) are learned. The feature space represents the features used for model training, that is, which features of the objects are learned.

[0090] Features can be understood as attributes or variables of data, or as characteristics corresponding to the data. Features can be obtained by preprocessing the raw data to enable the model to better understand and predict. Preprocessing can include one or more of sorting, vector representation, matrix representation, statistical averaging, feature extraction, or dimensionality reduction, and there is no limitation to this. In model training, labels correspond to features. Features describe the attributes or variables of the data, while labels are the target variables that the model wants to predict or classify. Let's use the example of training a model for image classification to illustrate. If a model is to be trained to distinguish between images of cats and dogs, a set of labeled images serves as training data. Each image consists of two parts: features and labels. In this example, features are attributes that describe the image and may include at least one of pixel values, color histograms, or texture features. These features provide key information about the image, helping the model understand its content. The label, on the other hand, is the result of classifying the image, i.e., whether the image is a cat or a dog.

[0091] For example, please refer to Table 1 below, which is an example of data.

[0092] Table 1

[0093] As shown in Table 1 above, the sample space corresponding to the data includes user 1 and user 2, and the feature space includes wireless access type, reference signal received power, reference signal received quality, and wireless access network throughput.

[0094] 2. Distributed Learning

[0095] Distributed learning refers to a learning method that combines data from multiple participants to learn a model. In the embodiments of this application, the model trained using distributed learning is referred to as a distributed learning model. The distributed learning model can be an AI model, an ML model, a feature, or an algorithm, etc., without specific limitation.

[0096] A typical example of distributed learning is federated learning (FL). Federated learning is a machine learning framework that can effectively help multiple users use data and conduct machine learning modeling while meeting the requirements of user privacy protection, data security, and government regulations. As a distributed machine learning paradigm, federated learning can effectively solve the data silo problem and conduct joint modeling without sharing user data, thereby technically breaking down data silos and achieving artificial intelligence (AI) collaboration. Distributed learning can be applied when two training entities cannot share their models (for example, due to privacy between the AF and NWDAF, or due to privacy between multi-vendor NWDAFs), and cannot share the input data used to train the model (to avoid large-scale data collection or privacy issues).

[0097] Federated learning can be categorized into three types based on the characteristics of the data sources of the participating parties: horizontal federated learning, transfer learning, and vertical federated learning (VFL). In horizontal federated learning, the sample space corresponding to the data has low overlap (or can be described as less overlap), but the sample features have high overlap (or can be described as more overlap). For example, the participants in horizontal federated learning include participants 1 and 2. The data provided by participant 1 includes user 1's features 1 and 2, and the data provided by participant 2 includes user 2's features 1 and 2. Transfer learning is a method that applies a model from a source task in a source domain to a target task in a target domain. For example, Task A (source task) is to classify cats and tigers in images, and Task B (target task) is to distinguish the lengths of cats and tigers in images. Traditional machine learning methods can only train two separate models for these two completely different tasks. However, with transfer learning, the model for Task A can be fully utilized and fine-tuned to obtain a model suitable for Task B.

[0098] As a machine learning technique, vertical federated learning can be used to address model training and inference challenges when participants are unwilling to share their original datasets. It is suitable for scenarios where the training sample spaces (e.g., samples) of participants (or participants) have high overlap, while the data (or features) of the participants have low overlap. Vertical federated learning combines different datasets (or features) of common samples from multiple participants for federated learning. This means that the data (or features) of each participant are partitioned vertically, hence the name vertical federated learning.

[0099] For example, please refer to Figure 1, which shows a data distribution diagram for a vertical federated learning scenario. Figure 1 takes the federated learning scenario as an example, with participants A and B as the participants. As shown in Figure 1, the sample spaces corresponding to participants A and B have a high degree of overlap, but the feature spaces have a low degree of overlap. For example, participant A's dataset contains data for user 1, user 2, user 3, and user 4, where each user's data includes features 1, 2, 3, 4, and 5. Participant B's dataset contains data for users 1, 2, 3, 4, and 5, where each user's data includes features 6, 7, 8, 9, and 10. This shows that the feature overlap between participants A and B is small, but the samples are mostly identical. When conducting vertical federated learning, participant A and participant B can select the features (i.e., feature 1, feature 2, feature 3, feature 4, feature 5, feature 6, feature 7, feature 8, feature 9, and feature 10) corresponding to the same samples (i.e., user 1, user 2, user 3, and user 4) from their respective data sets, as well as the labels corresponding to these features, to conduct model training for vertical federated learning.

[0100] All parties involved in distributed learning can be regarded as participants (or participants), and the participants can be divided into master participants and slave participants. There can be one or more slave participants, and there is no limitation on this. The master participant has the labels of the data required for the distributed learning task, and optionally has some of the data features required for the distributed learning task. The slave participant has some or all of the data features required for the vertical federated learning task. The master participant can also be called an active participant, an active participant (active participant) or an active client (active client), etc. The slave participant can also be called a passive participant (passive participant) or a passive client (passive client), etc.

[0101] Distributed learning also includes coordinators, or servers, etc. A coordinator, also known as a trusted coordinator, is responsible for maintaining the distributed learning process, authorizing the entry and removal of distributed learning members, and optionally distributing encryption keys and decrypting intermediate information. A coordinator is typically a third party independent of the task or a credible institution in the industry. Furthermore, the participant who initiates distributed learning can be called an initiator or active participant, while other participants can be considered passive participants. The initiator can be a master participant, a slave participant, or a coordinator in distributed learning, without limitation.

[0102] The main feature of distributed learning model training (such as VFL joint model training) is that there are at least two models (such as ML models), and each model is associated with a different training entity (such as a network element) (i.e., a master participant and at least one slave participant). The master participant trains their own models together with the models owned by the slave participants, where the two models have the same model objectives (for example, they are being trained to predict the same output, such as service experience), but the two models are different from each other (for example, they have different input data types, etc.). The optional participants participating in the distributed learning also include a distributed learning model training ID, which represents a unique identifier of the distributed learning process, which is used to identify the unique identifiers of at least two models (i.e., models from active and passive participants) and associate them with the distributed learning model training process.

[0103] Please refer to FIG2, which is a schematic diagram of a distributed learning scenario applicable to the embodiment of the present application. As shown in FIG2, the scenario includes a master participant, a slave participant, and a collaborator, and any two of the master participant, the slave participant, and the collaborator can communicate through a communication network, such as the fifth generation (5G) thgeneration, 5G) communication network, sixth generation communication network or future evolution communication network, etc.

[0104] Any participant in distributed learning (such as a master participant, a slave participant, or a collaborator) can be implemented through a network element. The network element involved in the embodiments of the present application can be implemented through a device, a client, software in a device (such as an application), a hardware module in a device (such as a chip), or a device cluster. The client can be, for example, an application running in a terminal device, or a terminal device running an application, etc. The client can also be, for example, an application or function in a network element, etc., and there is no limitation on this. For example, the network element is specifically a terminal device, a network device, or a third-party server, etc., and there is no limitation on this. The master participant has a distributed learning model of the master participant (or a local distributed learning model of the master participant), and the slave participant has a distributed learning model of the slave participant (or a local distributed learning model of the slave participant).

[0105] Figure 2 illustrates an example where the master is client 1, the slave is client 2, and the collaborator is the server. However, there are no practical limitations on how the master, slave, and collaborators can be implemented. Either the master or the slave can initiate distributed learning model training, and the collaborator can coordinate the distributed learning model training between the two parties.

[0106] Among them, the terminal device is a device with wireless transceiver functions, which can be a fixed device, mobile device, handheld device, wearable device, vehicle-mounted device, or a wireless device built into the above devices (for example, a communication module or chip system, etc.). The terminal device is used to connect people, objects, machines, etc., and can be widely used in various scenarios, such as but not limited to the following scenarios: cellular communication, device-to-device communication (device-to-device, D2D), vehicle to everything (vehicle to everything, V2X), machine-to-machine / machine-type communication (machine-to-machine / machine-type communication, M2M / MTC), Internet of Things (IoT), virtual reality (virtual reality, VR), augmented reality (augmented reality, AR), industrial control, self-driving, remote medical, smart grid, smart furniture, smart office, smart wearable, smart transportation, smart city, drones, robots and other scenarios. A terminal device may sometimes be referred to as user equipment (UE), terminal, access station, UE station, remote station, wireless communication device, or user equipment.

[0107] The network equipment includes, for example, access network equipment (or, referred to as access network devices / access network network elements), and / or core network equipment (or, referred to as core network devices / core network network elements). The access network equipment is a device with wireless transceiver functions, which is used to communicate with the terminal equipment. Access network equipment includes but is not limited to base stations (BTS, Node B, eNodeB / eNB, or gNodeB / gNB) in the communication system, transmission reception points (TRP), base stations of subsequent evolution of 3GPP, access nodes in wireless fidelity (WiFi) systems, wireless relay nodes, wireless backhaul nodes, satellites or drones, etc. The base station can be: a macro base station, a micro base station, a pico base station, a small station, a relay station, etc. Multiple base stations can support the network of the same access technology mentioned above, or they can support the network of different access technologies mentioned above. The base station can include one or more co-sited or non-co-sited transmission and receiving points. The access network device can also be a wireless controller, a centralized unit (CU), also called an aggregation unit, and / or a distributed unit (DU) in a cloud radio access network (C(R)AN) scenario. The access network device can also be a server, a wearable device, or a vehicle-mounted device. For example, the access network device in the vehicle to everything (V2X) technology can be a road side unit (RSU). The following describes the access network device by taking a base station as an example. The multiple access network devices in the communication system can be base stations of the same type or different types. The base station can communicate with the terminal device or communicate with the terminal device through a relay station. The terminal device can communicate with multiple base stations in different access technologies.

[0108] In the case where the access network device includes a CU and / or a DU. CU and DU can be understood as a division of the access network device from a logical functional perspective. CU and DU can be physically separated or deployed together, and this embodiment of the present application does not specifically limit this. A CU can be connected to a DU, or multiple DUs can share a CU. The division of CU and DU can be based on the protocol stack. One possible way is to deploy the radio resource control (RRC), service data adaptation protocol stack (SDAP) and packet data convergence protocol (PDCP) layers in the CU, and the remaining radio link control (RLC) layers, media access control (MAC) layers and physical layers in the DU. The embodiment of the present application does not completely limit the division of CU and DU according to the above-mentioned protocol stack method, and there may be other division methods, such as division according to service type.

[0109] Access network equipment may also refer to a centralized unit control plane (CU-CP) node or a centralized unit user plane (CU-UP) node, or include both the CU-CP and the CU-UP. The CU-CP is responsible for control plane functions, primarily including RRC and PDCP-C. PDCP-C is primarily responsible for encryption and decryption, integrity protection, and data transmission of control plane data. The CU-UP is responsible for user plane functions, primarily including SDAP and PDCP-U. SDAP is primarily responsible for processing core network data and mapping flows to bearers. PDCP-U is primarily responsible for encryption and decryption, integrity protection, header compression, sequence number maintenance, and data transmission of the data plane.

[0110] In different systems, CU (including CU-CP or CU-UP) or DU may have different names, but those skilled in the art will understand their meanings. For example, in an open radio access network (O-RAN) system, CU may also be referred to as O-CU (Open CU), DU may also be referred to as O-DU, CU-CP may also be referred to as O-CU-CP, and CU-UP may also be referred to as O-CU-UP.

[0111] The core network equipment is used to implement at least one of the following functions: mobility management, data processing, session management, policy and billing. The names of the devices that implement the core network functions in systems with different access technologies may be different, and this embodiment of the present application is not limited to this. Taking the 5G system as an example, the core network equipment includes: access and mobility management function (AMF), session management function (SMF), or user plane function (UPF).

[0112] The following describes the basic process of distributed learning model training, combined with the process diagram of vertical federated learning shown in Figure 3. Figure 3 takes the distributed learning using the homomorphic encryption algorithm as an example. The master participant involved in Figure 3 is, for example, the master participant involved in Figure 2, the slave participant involved in Figure 3 is, for example, the slave participant involved in Figure 2, and the collaborator involved in Figure 3 is, for example, the collaborator involved in Figure 2.

[0113] S301. Both the slave participant and the master participant initialize local parameters.

[0114] Initialize the local parameters Θ of the local distributed learning model from the participants A , the master party initializes the local parameters Θ of the master party's local distributed learning model B .

[0115] S302: The collaborating party creates an encryption key pair. The collaborating party creates a homomorphic encryption key pair based on a homomorphic encryption algorithm.

[0116] S303: The collaborating party sends the public key to the slave party and the master party.

[0117] Specifically, S303 includes S303a and S303b, wherein S303a is the collaboration party sending the public key to the master party, and the master party receives the public key from the collaboration party. S303b is the collaboration party sending the public key to the slave party, and the slave party receives the public key from the collaboration party.

[0118] S304: Determine local parameters and loss functions from the participants.

[0119] The slave party can train the local distributed learning model of the slave party based on the data of the slave party, and update the local parameters and loss function of the local distributed learning model, and encrypt the local parameters and loss function based on the public key to obtain the encrypted local parameters, for example The encrypted loss function is, for example, in The symbol represents homomorphic encryption.

[0120] S305: The slave participant sends the encrypted local parameters and loss function to the master participant. Correspondingly, the master participant receives the encrypted local parameters and loss function from the slave participant.

[0121] For example, local parameters from the participants and loss function Sent to the main participant.

[0122] S306: The main participant determines local parameters, intermediate parameters and loss function.

[0123] Similarly, the main participant can train the local distributed learning model of the main participant based on the data of the main participant, update the local parameters, and encrypt the local parameters based on the public key to obtain the encrypted local parameters, for example And the encrypted loss function is The master party can also determine the intermediate parameters based on the master party's local parameters and the slave party's local parameters. Intermediate parameters For example it could be The master party can also use the local parameters of the slave party and loss function And the loss function etc., determine the total loss function

[0124] S307: The master participant sends the intermediate parameters to the slave participant. Correspondingly, the slave participant receives the intermediate parameters from the master participant.

[0125] For example, the main participant will Issued to the participating parties.

[0126] S308: The main participant sends the loss function to the collaborating party. Correspondingly, the collaborating party receives the loss function from the main participant.

[0127] For example, the main participant will lose the function Sent to collaborators.

[0128] S309: The collaborator determines whether the iteration is terminated.

[0129] For example, the collaborator uses the private key to decrypt the loss function Obtain the decrypted loss function, and determine whether the iteration is terminated based on the decrypted loss function, or it can be described as determining whether to stop the distributed learning model training.

[0130] S310: Determine the encrypted local gradient from the participant.

[0131] Compute local gradients from participating parties And add random noise Get the encrypted local gradient, that is

[0132] S311: The slave participant sends the encrypted local gradient to the collaboration partner. Correspondingly, the collaboration partner receives the encrypted local gradient from the slave participant.

[0133] The collaborator uses the private key to decrypt the encrypted local gradient and obtains the decrypted result, i.e.

[0134] S312: The collaboration direction sends the decryption result to the slave participant, and correspondingly, the slave participant receives the decryption result from the collaboration direction.

[0135] Subtract random noise from the decrypted result to get the decrypted gradient And update the local parameters based on the decrypted gradient

[0136] S313. The main participant determines the encrypted local gradient.

[0137] The master party calculates the local gradient And add random noise Get the encrypted local gradient, that is

[0138] S314: The active participant sends the encrypted local gradient to the collaborating party. Correspondingly, the collaborating party receives the encrypted local gradient from the active participant.

[0139] The collaborator uses the private key to decrypt the encrypted local gradient and obtains the decrypted result, i.e.

[0140] S315: The collaborating party sends the decryption result to the main party. Correspondingly, the main party receives the decryption result from the collaborating party.

[0141] Similarly, the main participant subtracts random noise from the decrypted result to obtain the decrypted gradient And update the local parameters based on the decrypted gradient

[0142] During the training process, participants in distributed learning exchange intermediate computational results. Therefore, none of the participants can access each other's original data. Instead, they complete the model training process by exchanging intermediate results. This not only enables joint training but also protects the privacy of the participants' local data.

[0143] Figure 3 uses two participants as an example. When there are three or more participants, the distributed learning process involved can be referred to in the distributed learning process described in Figure 3 and will not be listed here. Distributed learning can involve a variety of encryption algorithms or model training methods, such as multi-party multi-classification distributed learning based on privacy-preserving label sharing and multi-party distributed learning based on secret sharing, without specific limitation.

[0144] Because a participant may not possess all the features required for distributed learning, it is necessary to collaborate with other participants to jointly train the distributed learning model. This requires identifying the other participants in the vertical federation. This means that before model training begins, the features of each participant must be aligned. This means determining the features that each participant in the vertical federation will use for distributed learning model training. However, currently, negotiation with each participant is done offline, resulting in limited flexibility in determining participants and failing to meet certain business needs.

[0145] In view of this, an embodiment of the present application provides a communication method. In this method, each network element can be registered in the second network element. In this way, the initiator of a distributed learning (such as the first network element) can obtain some network elements (such as at least one third network element) that meet the requirements of distributed learning through the second network element. In this way, the first network element can determine the participants who will ultimately participate in the distributed learning from the at least one third network element obtained. There is no need to determine the participants in advance, thereby realizing a mechanism for online determination of distributed participants, which is conducive to improving the flexibility of determining participants. In addition, the first network element can also determine the features of the participants used to participate in the training of the distributed learning model. In this way, a method for aligning the features of each participant is provided, which is also conducive to improving the flexibility of aligning the features of each participant.

[0146] The communication method provided in the embodiment of the present application can be applied to the distributed learning scenario involved in Figure 2 above. In addition, it can also be applied to other similar distributed learning scenarios, which are introduced below with examples in conjunction with the accompanying drawings.

[0147] Please refer to Figure 4, which is a schematic diagram of a distributed learning scenario provided in an embodiment of the present application. Figure 4 illustrates an initiator, other participants, and a registrant. Optionally, the scenario also includes collaborators. Any of the initiator, other participants, registrant, and collaborator can be implemented through a network element. The other participants can be one or more, without limitation. Optionally, the registrant can also be a collaborator in distributed learning.

[0148] Figure 4 takes the example of a first network element as the initiator, a second network element as the registrant, some or all of at least one third network element as the other participants, and a fourth network element as the collaborator. The initiator (e.g., the first network element) initiates distributed learning. The registrant (e.g., the second network element) registers the initiator and other participants. The collaborator can assist in training the distributed learning model, etc. The implementation of the network element can refer to the network element content discussed above, and the repeated parts are not listed here.

[0149] Exemplarily, at least one third network element and the first network element may both be registered in the second network element. When the first network element initiates distributed learning, it may request the second network element to participate in the distributed third network element.

[0150] The 3GPP standard defines an intelligent network architecture based on the Network Data Analytics Function (NWDAF). This intelligent network architecture aims to collect massive amounts of information from the network and utilize big data and artificial intelligence technologies (such as vertical federated learning) to leverage this data. This output provides valuable information to assist operators in policy formulation and network resource adjustments, thereby improving user experience and reducing network load.

[0151] Please refer to Figure 5, which is a schematic diagram of an intelligent network architecture based on NWDAF provided in an embodiment of the present application, which can also be regarded as a schematic diagram of another distributed learning scenario provided in an embodiment of the present application. Figure 5 illustrates NWDAF, a vertical federated learning server (VFL Server), a vertical federated learning support function (VFLSF), an application function (AF), a network function (NF), a UE, and a RAN. NF includes, for example, a network exposure function (NEF), a network repository function (NRF), a network data analytics function (NWDAF), a policy control function (PCF), an access and mobility management function (AMF), a session management function (SMF), operations, administration, and management (OAM) (also known as network management), and a user plane function (UPF).

[0152] Any one of the NF, UE, or RAN involved in FIG5 can be used as an example of the first network element involved in FIG4 , at least one of the NF, UE, or RAN involved in FIG5 can be used as an example of at least one third network element involved in FIG4 , and the NRF, NEF, or VFLSF involved in FIG5 can be used as an example of a second network element involved in FIG4 . In addition, optionally, the VFL Server involved in FIG5 can serve as a collaborator of distributed learning. For example, the VFLSF can be a newly added network element, or can be deployed in an existing network element, for example, in a network element such as an NRF, NEF, NWDAF, or AF.

[0153] NWDAF has functions such as data collection, model training, data analysis, and model inference. It can be used to collect relevant data from other network elements, third-party servers, terminal devices, or network management systems, perform data analysis or model training based on the relevant data, and provide data analysis results to network elements, third-party service servers, terminal devices, or network management systems, or provide trained models to other network elements with data analysis functions. NWDAF can be divided into analytics logical function (AnLF) and model training logical function (MTLF) based on its functions. AnLF is a logical function in NWDAF, used to perform model inference, derive analysis results (i.e., derive statistical or predictive analysis results based on the analysis consumer's request), and make analysis results available. MTLF is a logical function in NWDAF, used to train models and make training services available (for example, providing trained models). An NWDAF may contain only AnLF, only MTLF, or both AnLF and MTLF.

[0154] Please refer to Table 2 below for the analysis results provided by NWDAF and the data that NWDAF needs to collect to provide the corresponding analysis results.

[0155] Table 2

[0156] As shown in Table 2 above, NWDAF can provide service experience analysis results. To provide this analysis result, NWDAF needs to collect service-related data such as service identification and service experience from AF / UE, as well as data such as signal reception power and signal reception quality from OAM. NWDAF can train AI models based on the collected data, and then obtain inferred analysis results based on the AI ​​models, such as the predicted service experience for a certain time period in the future. NWDAF can provide network element load analysis results. To provide this analysis result, NWDAF needs to collect information related to network element resource usage from NRF, as well as information such as UE speed / direction, and data such as UE path, destination, and arrival time from third-party applications. NWDAF can train AI models based on the collected data, and obtain inferred analysis results based on the AI ​​models, such as the predicted network element load for a certain time period in the future. NWDAF can also provide MEC service experience analysis. To provide this analysis result, NWDAF needs to collect UE identification, UE location, application identification, application location, and uplink / downlink transmission performance data from AF or UE, train an AI model based on this data, and obtain inference analysis results based on the AI ​​model, such as a predicted MEC service experience analysis for a certain time period in the future.

[0157] The AF is used to convey the requirements of the application side to the network side, such as quality of service (QoS) requirements or user status event subscriptions, etc. The AF can be a third-party functional entity or an application server deployed by the operator.

[0158] VFLSF is responsible for the registration and discovery of participants in vertical federated learning. VFLSF can be deployed independently or in network elements such as NRF, NEF, NWDAF, or AF.

[0159] For example, the VFL Server can act as a collaborator, responsible for maintaining the vertical federation process, authorizing the admission and removal of federated learning members, etc. Optionally, the VFL Server is responsible for tasks such as distributing encryption keys and decrypting intermediate information.

[0160] NEF is used to provide the framework, authentication, and interface related to network capability exposure, and to transmit information between 5G system network functions and other network functions.

[0161] NRF is used to provide registration and discovery capabilities for network elements in the network.

[0162] PCF is mainly responsible for the generation and update of UE access strategy and QoS flow control strategy.

[0163] The AMF is mainly responsible for user access and mobility management functions, including user registration, reachability, mobility management, N1 / N2 interface signaling transmission, access authentication and authorization, etc.

[0164] SMF is used to manage the creation, modification, and release of user protocol data unit (PDU) sessions, as well as the allocation and management of IP addresses, the selection and control of UPFs, etc.

[0165] OAM is used to provide system or network fault indication, performance monitoring, security management, diagnosis, configuration and user configuration.

[0166] As a network function element of the 5G core network, UPF undertakes core network processing functions such as data traffic processing and routing forwarding.

[0167] In Figure 5, Nnef, Nnrf, Nnwdaf, Naf, Npcf, Nvflsf, Namf, Nsmf, and Nvfl-server are service-oriented interfaces provided by the NEF, NRF, NWDAF, AF, PCF, VFLSF, AMF, SFM, and VFL-Server, respectively, for invoking corresponding service-oriented operations. N2 in Figure 5 is the communication interface between the RAN and the core network control plane (e.g., AMF). N3 is the communication interface between the RAN and the UPF, used to transmit user data. N4 is the communication interface between the SMF and the UPF, used for policy configuration, etc., for the UPF.

[0168] Please refer to Figure 6, which is a schematic diagram of a distributed learning scenario provided in an embodiment of the present application. Figure 6 takes the first network element as AF, the second network element as NRF, and the third network element as NWDAF as an example. In fact, the specific implementation methods of the first network element, the second network element, and the third network element are not limited. In this case, AF can serve as the initiator of distributed learning model training, NRF can serve as the registrant of distributed learning model training, and NWDAF can serve as a participant in distributed learning model training.

[0169] The communication method provided in the embodiments of the present application is introduced below with reference to the accompanying drawings.

[0170] In the various embodiments of the present application, the steps indicated by dotted lines are optional steps. In addition, the first network element involved in the various embodiments of the present application is, for example, the master participant or slave participant involved in Figure 2, the initiator or first network element involved in Figure 4, or the first network element or AF involved in Figure 6, the second network element is, for example, the registrant or second network element involved in Figure 4, or the VFLSF or NRF or NEF involved in Figure 5, or the second network element or NRF involved in Figure 6, the third network element is, for example, the master participant or slave participant involved in Figure 2, the other participants or any third network element involved in Figure 4, or any network element involved in Figure 5, or the third network element or NWDAF involved in Figure 6, and the collaborator is, for example, the collaborator involved in Figure 2, or the collaborator or fourth network element involved in Figure 4. These network elements may have other names or the names of such network elements may change as the standards continue to evolve, and there is no limitation on this.

[0171] FIG7 illustrates a communication method provided in an embodiment of the present application. The steps involved in FIG7 are introduced below.

[0172] S701: A first network element sends a first request to a second network element. Correspondingly, the second network element receives the first request from the first network element.

[0173] The first network element may send the first request directly to the second network element, or the second network element may send the first request to the second network element through other network elements (which may be one or more network elements), without limitation. For example, if the first network element is an AF and the second network element is an NRF, the AF may send the first request to the NRF through the NEF. Alternatively, if the first network element is an NWDAF and the second network element is an NRF, the NWDAF may send the first request directly to the NRF. Alternatively, if the first network element is an NWDAF and the second network element is an AF, the NWDAF may send the first request to the AF through the NEF.

[0174] The first network element can obtain a task, which may be obtained from a terminal device corresponding to the user, or obtained from other network elements, without limitation. Based on the requirements of the task, the first network element can determine that a model for processing the task (called a distributed learning model) needs to be trained in a distributed learning manner, and determine multiple features for training the distributed learning model. For example, the first network element can determine multiple features corresponding to the task in the first correspondence based on the first correspondence. The first correspondence indicates features corresponding to different tasks. Alternatively, the first network element can determine features corresponding to multiple data corresponding to the task in the second correspondence based on the second correspondence. The first correspondence indicates data corresponding to different tasks.

[0175] For example, a first network element obtains a task for predicting service experience. A first correspondence indicates that the features corresponding to the task for predicting service experience include RAT type, RSRP, RSRQ, RAN throughput, service experience, location, average throughput, and average packet delay. Based on the task, the first network element can determine multiple features including RAT type, RSRP, RSRQ, RAN throughput, service experience, location, average throughput, and average packet delay.

[0176] Alternatively, the first network element may also obtain multiple features for training the distributed learning model from other network elements. These multiple features may be understood as features corresponding to the data required for training the distributed learning model, or may be understood as the features required to train the distributed learning model.

[0177] After the first network element determines multiple features, the first network element may determine the multiple features except the features that the first network element can provide (or support) as one or more features. The features that the first network element can provide (or support) may refer to features that the first network element can directly obtain, and may also include features corresponding to data that the first network element can collect from other network elements. These one or more features are used for distributed learning model training, and these one or more features can also be understood as the features expected by the first network element (expected all features), or can be understood as all features that the first network element expects other participants to provide for distributed learning model training.

[0178] Since the first network element cannot support the one or more features, it is necessary to discover a network element for providing the one or more features so that the first network element and the discovered network element can jointly perform distributed learning model training. In an embodiment of the present application, the first network element may send a first request to the second network element for requesting the discovery of the network element. The number of network elements requested for discovery may be one or more, and there is no limitation on this. The first request indicates (or includes) one or more features. The first request is used to request the discovered network element to participate in the distributed learning model training. The features include the one or more features.

[0179] The first request may include K bits, where K is a positive integer, and these K bits are used to request the discovery of network elements, that is, the first request uses a separate bit to request the discovery of network elements. Alternatively, the second network element determines that the first request is used to request the discovery of network elements based on the name of the first request or the information of the service operation corresponding to the first request (such as the name). For example, the second network element is NRF. If the first request is the Nnrf_NF Discovery_Request service operation to send the first request to NRF, then NRF determines that the first network element is to discover network elements based on the name of the service operation. In this case, there is no need for the first request to use a separate bit to indicate the discovery of network elements, which helps to save the number of bits occupied by the first request. Alternatively, the one or more features are used to request the discovery of network elements. In this case, the first request does not need to use a separate bit to request the discovery of network elements, saving the bit overhead of the first request. In other words, the first request indicates one or more features, which means that the first request is used to request the discovery of network elements.

[0180] The first request may indicate any one of the following items A1 to A5 to indicate one or more features. The content of any one of items A1 to A5 is introduced below.

[0181] A1. The first request includes one or more characteristics, such as at least one of location, average throughput, or average packet delay.

[0182] A2. The first request includes one or more feature identifiers (feature IDs), where the one or more feature identifiers are used to indicate the one or more features.

[0183] The identifier of a feature (such as the identifier of one of one or more features) can be an identifier assigned to the feature (which can also be called a number, serial number or index, etc.). In one possible implementation, the identifier of one of the one or more features (which can be simply referred to as a feature identifier) ​​can be understood as an anonymous feature, and the participants (such as the first network element and other participants) negotiate the meaning of the feature identifier in advance, so that they can understand each other's registered feature identifiers. Optionally, the second network element cannot know which feature the feature identifier specifically represents through the feature identifier, so as to avoid leaking the features supported by each participant to other parties other than the participants in the distributed learning. In other words, the feature identifier can be understood by the participants in the distributed learning model, but not by the second network element, thus ensuring the security of the feature. In another possible implementation, the feature identifier can also be non-anonymous, and the participants do not need to negotiate the feature identifier in advance. The embodiment of the present application does not need to limit this.

[0184] For example, a feature identifier of 1234 indicates that the feature is average throughput; or a feature identifier of 1235 indicates that the feature is average packet delay.

[0185] A3. The first request includes one or more event identifiers (Event IDs). Each event identifier is used to determine data supporting model training (a list of Event IDs of the local data for training). The event identifiers involved in the various embodiments of the present application can be preconfigured or predefined, for example, predefined by a protocol, or can be determined by negotiation between the first network element and at least one third network element, etc., without limitation.

[0186] An event identifier may correspond to (or may determine) at least one feature. Accordingly, one of the one or more event identifiers may correspond to some or all of the one or more features. For example, if an event identifier is QoS Monitoring, then the feature corresponding to the event identifier may include at least one of uplink packet delay (UL packet delay), downlink packet delay (DL packet delay), or round trip packet delay.

[0187] For example, the first request includes {event ID1, event ID2, event ID4, event ID5, feature ID3, feature ID4}, that is, the one or more features include at least one feature corresponding to event 1, at least one feature corresponding to event 2, at least one feature corresponding to event 4, at least one feature corresponding to event 5, feature 3 and feature 4.

[0188] A4. The first request includes one or more network element types (NF type(s)) and / or one or more network element set identifiers (NF Set ID(s)). The first request may indicate one or more characteristics through one or more network element types or one or more network element set identifiers.

[0189] One or more network element types or one or more network element sets represent data sources that can be collected as data. Data corresponds to characteristics, so indicating one or more network element types or one or more network element sets is equivalent to indicating one or more characteristics.

[0190] For example, an NF type or an NF Set ID can correspond to one or more Event IDs, and an Event ID can correspond to one or more features. For example, if NF type = SMF, the Event IDs that SMF can provide include, for example, QoS flow monitoring, uplink path change, and QoS flow identifier (QFI) allocation. Features corresponding to QoS flow monitoring include, for example, uplink packet delay, downlink data delay, or round-trip packet delay.

[0191] In one possible implementation, in addition to indicating one or more features, the first request may also indicate one or more network element types (NF type(s)). For example, the first request includes one or more network element type identifiers, and the one or more network element type identifiers indicate the one or more network element types. Under this implementation, each of the one or more features is a feature corresponding to one of the one or more network element types. Of course, in the case where one or more features are indicated by one or more network element types in A4 above, the first request can indicate both one or more features and one or more network element types by indicating the one or more network types.

[0192] In the case where the first request indicates one or more features and one or more network element types, the first request can be alternatively described as indicating one or more network element types and features corresponding to each of the one or more network element types. This can be understood as the first network element expecting the one or more features to be obtained from certain specific network element types, or as the first network element requesting one or more features at the granularity of network element type.

[0193] For example, the first request indication {NWDAF: (Event ID 1, Event ID 2, Event ID 4, Event ID 5), AF: (Feature ID3, Feature ID4)} indicates that the first network element expects to discover one or more NWDAF supported features (Event ID 1, Event ID 2, Event ID 4, Event ID 5), and at the same time, the first network element expects to discover one or more AF supported features {Feature ID3, Feature ID4}.

[0194] A5. The first request includes one or more capability types, each of which corresponds to at least one feature. Thus, the one or more capability types indicate one or more features.

[0195] For example, the first request includes active participation (active participant), and the characteristics corresponding to the capability are characteristic 1 and characteristic 2, indicating that one or more characteristics include characteristic 1 and characteristic 2.

[0196] In one possible implementation, in addition to indicating one or more characteristics, the first request may also indicate a VFL capability type. This situation applies to the case where the first request indicates one or more characteristics through any of the above methods such as A1-A4. Optionally, in this case, the first request may indicate one or more capability types, and the characteristics corresponding to each capability type. Of course, in the case where one or more characteristics are through one or more network element types in A4 above, the first request may indicate both one or more characteristics and one or more network element types by indicating the one or more network element types.

[0197] For example, if the capability type indicated in the first request includes active participation and the features corresponding to the active participation are feature 3 and feature 4, then it means that the one or more features include feature 3 and feature 4.

[0198] The above A1 or A2 can be regarded as the first request indicating the fine-grained features requested to be discovered, that is, one or more features in this case can be regarded as fine-grained features, and the above A3 to A5 can be regarded as the first request indicating the coarse-grained features requested to be discovered, that is, one or more features in this case can be regarded as coarse-grained features.

[0199] S702: The second network element sends first information to the first network element. Accordingly, the second network element receives the first information from the first network element. The first information indicates at least one third network element, where M third network elements participate in the distributed learning model training. M is a positive integer, such as 1, 2, or 3, and is not limited thereto.

[0200] After receiving the first request, the second network element may determine (or perceive or discover) at least one third network element based on the first request. The second network element may be preconfigured or predefined with information about the at least one third network element, or may obtain the information about the at least one third network element from the at least one third network element, without specific limitation. Of course, in addition to determining the at least one third network element, the second network element may also determine other network elements, such as the first network element, without limitation.

[0201] The information about a third network element in the information about at least one third network element may include an identifier or address of the third network element. Optionally, the information about a third network element may also include a network element type and / or network element capabilities of the third network element, which are not specifically limited. The network element capabilities of a third network element may, for example, indicate that the third network element has at least one of distributed learning capabilities and distributed learning model training or inference capabilities. Distributed learning capabilities may, for example, be vertical federated learning capabilities. Vertical federated learning capabilities include supported vertical federated learning capability types, such as vertical federated learning master participants, vertical federated learning slave participants, vertical federated learning collaborators, or time windows supported by vertical federated learning. Time windows include, for example, time periods for vertical federated learning, time periods for collecting data for vertical federated learning, or time periods for generating data for vertical federated learning. The identifier of a third network element may be the address of the third network element, an identifier assigned to the third network element, a computation performed on the address of the third network element, or the name of the third network element. The address of the third network element is, for example, an Internet protocol (IP) address of the third network element or a fully-qualified domain name (FQDN) of the third network element.

[0202] Because the second network element is able to determine at least one third network element, the second network element can indicate the at least one third network element to the first network element. For example, the second network element can send first information to the first network element, where the first information indicates the at least one third network element. The first information can also be considered a response message to the first request. The at least one third network element can be understood as a third network element that satisfies the first request, or can be understood as a network element discovered by the second network element in accordance with the first request.

[0203] Exemplarily, the first information includes information of at least one third network element, for example, an identifier of at least one third network element, and the information of at least one third network element is equivalent to indicating at least one third network element.

[0204] In addition to indicating the at least one third network element, the first information may optionally further indicate features corresponding to the at least one third network element. The features corresponding to the at least one third network element include features corresponding to each of the at least one third network element. The features corresponding to the at least one third network element may all be features among the one or more features indicated in the first request, or the features corresponding to the at least one third network element may include one or more features, as well as features other than the one or more features. For example, the features corresponding to the at least one third network element include features supported by each of the at least one third network element. This is not specifically limited.

[0205] The following takes the feature corresponding to one of the at least one third network element as an example to introduce the content of the feature corresponding to the at least one third network element.

[0206] B1. Features corresponding to a third network element may be features supported by the third network element, such as all features supported by the third network element. Features supported by the third network element may be, for example, all features that the third network element can provide.

[0207] For example, all features supported by the third network element include feature 1, feature 2, and feature 3, so the features corresponding to the third network element include feature 1, feature 2, and feature 3.

[0208] Under B1, the features corresponding to any two third network elements in at least one third network element may or may not overlap, and this is not specifically limited. The overlapping features corresponding to any two third network elements means that the features corresponding to at least two third network elements have the same features or are completely identical features. The non-overlapping features corresponding to any two third network elements means that the features corresponding to any two third network elements are different.

[0209] For example, at least one third network element includes a third network element A and a third network element B. The features corresponding to the third network element A include features 1, 2, and 3, and the features corresponding to the third network element B include features 1 and 4. Since feature 1 corresponding to the third network element A is the same as feature 1 corresponding to the third network element B, it can be considered that the features corresponding to the third network element A and the features corresponding to the third network element B overlap. Alternatively, at least one third network element includes a third network element C and a third network element D. The features corresponding to the third network element C include features 1 and 2, and the features corresponding to the third network element D include features 4 and 5. Since the features corresponding to the third network element C and the features corresponding to the third network element D are different, it can be considered that the features corresponding to the third network element C and the features corresponding to the third network element D do not overlap.

[0210] Under B1, the characteristics corresponding to at least one third network element include one or more characteristics indicated by the first request. Optionally, the characteristics corresponding to at least one third network element may also include other characteristics in addition to the one or more characteristics.

[0211] B2. The feature corresponding to a third network element is a feature supported by the third network element and belongs to one or more features indicated by the first request. In this case, the feature corresponding to a third network element is the intersection of the feature supported by the third network element and one or more features.

[0212] For example, one or more features include feature 1 and feature 2, and all features supported by the third network element include feature 1, feature 2, and feature 3. Then the features corresponding to the third network element are feature 1 and feature 2.

[0213] Under B2, features corresponding to any two third network elements in at least one third network element may or may not overlap, and there is no specific limitation on this.

[0214] Under B2, the characteristics corresponding to at least one third network element belong to one or more characteristics.

[0215] B3. A feature corresponding to a third network element is a feature used by the third network element to participate in distributed learning model training. The feature used by the third network element to participate in distributed learning model training can be understood as a feature determined by the second network element for the third network element to participate in distributed learning model training.

[0216] For example, one or more features include feature 1, feature 2, feature 3, and feature 4, and at least one third network element includes third network element A and third network element B. Third network element A supports features 1, 2, and 3, and third network element B supports features 3, 4, and 5. The second network element determines that third network element A uses features 1 and 2 for distributed learning model training, and that third network element B uses features 3 and 4 for distributed learning model training. Therefore, for third network element A, the second network element only sends features 1 and 2 to the first network element, or can be described as the second network element indicating features 1 and 2 corresponding to third network element A to the first network element. For third network element B, the second network element only sends features 3 and 4 to the first network element, or can be described as the second network element indicating features 3 and 4 corresponding to third network element B to the first network element. The union of the features of third network element A and third network element B constitutes one or more features, and there is no intersection between the features of third network element A and third network element B.

[0217] In this case, the feature corresponding to one of the third network elements is part or all of the intersection of the feature supported by the third network element and one or more features.

[0218] Under B3, features corresponding to any two third network elements in at least one third network element may or may not overlap, and there is no specific limitation on this.

[0219] Optionally, under B3, the features corresponding to each of the at least one third network element can be features from the one or more features, and the features corresponding to any two of the at least one third network element do not overlap. In this optional approach, the at least one third network element determined by the second network element for the first network element is equivalent to being a third network element that can be directly used to participate in the training of the distributed learning model, without the first network element having to further screen from the at least one third network element.

[0220] The following is an example of the specific content of the first information when the first information also indicates a feature corresponding to at least one third network element.

[0221] C1. The first information includes an identifier of at least one third network element and a characteristic corresponding to each of the at least one third network element. The characteristics corresponding to a third network element may refer to the characteristics corresponding to a third network element discussed above. In this way, the first network element can intuitively identify the characteristics corresponding to the at least one third network element.

[0222] For example, the first information includes {AF1: feature 1, feature 2 and feature 3, UE2: feature 4 and feature 5, AMF1: feature 1, feature 5 and feature 6}, then the at least one third network element indicated by the first information includes AF1, UE2 and AMF1, the characteristics corresponding to AF1 include feature 1, feature 2 and feature 3, the characteristics corresponding to UE2 include feature 4 and feature 5, and the characteristics corresponding to AMF2 include feature 1, feature 5 and feature 6.

[0223] C2. The first information includes information about at least one network element group. The information about each network element group in the at least one network element group includes identifiers of the network elements included in each network element group. The features corresponding to each network element group include one or more features. Optionally, the features corresponding to any two network elements in each network element group in the at least one network element group do not overlap. However, the features corresponding to any two network elements may overlap.

[0224] At least one network element group includes at least one third network element, and accordingly, one network element group within the at least one network element group includes some or all of the at least one third network element. In this manner, the second network element groups the at least one third network element, facilitating subsequent selection by the first network element of network elements for participating in distributed learning model training.

[0225] Optionally, the information of each network element group also includes characteristics corresponding to each network element group, and / or the information of each network element group also includes an identifier of each network element group. The characteristics corresponding to at least one network element group are characteristics corresponding to at least one third network element. And the characteristics corresponding to a network element group include one or more characteristics.

[0226] For example, the first information includes {Group 1: VFL client 1: Event ID 1, Event ID 4; VFL client 2: Event ID 2, Event ID 5; VFL client 3: Feature ID 3, Feature ID 4}, and {Group 2: VFL client 4: Event ID 1, Event ID 2, Event ID 4, Event ID 5; VFL client 5: Feature ID3; VFL client 6: Feature ID4}, then the first information indicates the information of two network element groups (i.e., network element group 1 and network element group 2).

[0227] Network element group 1 includes VFL client 1, VFL client 2, and VFL client 3. The features corresponding to VFL client 1 include the features corresponding to event 1 and the features corresponding to event 4. The features corresponding to VFL client 2 include the features corresponding to event 2 and the features corresponding to event 5. The features corresponding to VFL client 3 include features 3 and features 4. Therefore, the features corresponding to network element group 1 include the features corresponding to VFL client 1, the features corresponding to VFL client 2, and the features corresponding to VFL client 3. Network element group 2 includes VFL client 4, VFL client 5, and VFL client 6. The features corresponding to VFL client 4 include the features corresponding to event 1, the features corresponding to event 2, the features corresponding to event 4, and the features corresponding to event 5. The features corresponding to VFL client 5 include features 3 and features 4 corresponding to VFL client 6. Therefore, the features corresponding to network element group 2 include the features corresponding to VFL client 4, the features corresponding to VFL client 5, and the features corresponding to VFL client 6.

[0228] In one possible implementation, when the first request further indicates one or more network element types, the network element type of the at least one third network element may belong to one or more network element types, and the characteristics corresponding to each third network element in the at least one third network element may belong to one or more characteristics. If the first information is the information shown in C2 above, then the network element type corresponding to the at least one network element group belongs to one or more network element types. For example, the network element type corresponding to the at least one network element group belongs to one network element type among the one or more network element types, and this is not specifically limited.

[0229] In one possible design, after the second network element sends the first information to the first network element, the first network element can use at least one third network element as another participant in the distributed model learning and training. In this case, the M third network elements are all the network elements in the at least one third network element.

[0230] For example, if the first information does not indicate a feature corresponding to the at least one third network element, the first network element may further negotiate with each of the at least one third network element about a feature supported by each third network element.

[0231] Alternatively, if the first information also indicates a feature corresponding to at least one third network element, then the first network element may determine that the feature corresponding to the at least one third network element is a feature participating in distributed model learning and training.

[0232] Alternatively, if the first information also indicates features corresponding to at least one third network element, and the first request indicates the coarse-grained features requested to be discovered (for example, the first request indicates one or more features in the above-mentioned A3 to A5 manner (such as the one or more features indicated by the first request may be Event ID or NF type(s), NF Set ID(s) or capability type, etc.)), the first network element may further negotiate with each third network element in at least one third network element about the features supported by each third network element. Optionally, the first network element may negotiate with each third network element in at least one third network element about finer-grained features for distributed model training, etc.

[0233] In another possible implementation, the first network element determines M third network elements from at least one third network element based on the first information. In this case, the M third network elements may be all of the at least one third network element, or may be part of the at least one third network element, without specific limitation. These M third network elements can be understood as network elements determined by the first network element to participate in the distributed learning model training.

[0234] The methods of determining the M third network elements are different depending on the content of the first information, which are introduced below respectively.

[0235] D1. When the first information also indicates characteristics corresponding to at least one third network element, the first network element may directly determine the characteristics corresponding to the at least one third network element based on the first information.

[0236] When the first information indicates information of at least one third network element and the corresponding characteristics of each third network element in the at least one third network element, the first network element can determine from the at least one third network element M third network elements whose corresponding characteristics include one or more characteristics, that is, determine the M third network elements participating in the training of the distributed learning model.

[0237] Optionally, when the feature corresponding to one of the at least one third network elements is a feature used by the third network element to participate in the training of the distributed learning model, the first network element can directly determine at least one third network element as M third network elements, which can reduce the processing amount of the first network element.

[0238] Alternatively, the first network element can determine multiple candidate network element groups from at least one third network element, each candidate network element group includes part or all of the network elements in at least one third network element, and the characteristics corresponding to each candidate network element group include one or more characteristics, and the first network element can determine a candidate network element group from multiple candidate network element groups, and the network elements included in the determined candidate network element group are M third network elements.

[0239] The determined candidate network element group (i.e., the candidate network element group corresponding to the M third network elements) can be the network element group with the smallest sum of distances to the first network element, thereby facilitating shortening communication time. Alternatively, the determined candidate network element group can be a network element group with a relatively low overall load, thereby facilitating load balancing of the communication system. The determined candidate network element group can be the network element group with the best overall communication quality with the first network element, thereby facilitating improving the efficiency of distributed learning model training.

[0240] D2. When the first information includes information about at least one network element group, the first network element may directly select a network element group from the at least one network element group. The network elements included in the selected network element group are the M third network elements. The content of the selected network element group can be determined as a candidate group by referring to the above description and is not listed here one by one.

[0241] D3. When the first information only indicates at least one third network element (or it can be understood that the first information does not indicate the characteristics corresponding to at least one third network element), the first network element can request the characteristics corresponding to at least one third network element from at least one third network element, and then determine M third network elements based on the characteristics corresponding to at least one third network element.

[0242] For example, the first network element sends third information to at least one third network element. Similarly, the first network element sends at least one third information in total, where the third information sent to one of the at least one third network element is used to request the third network element to provide feedback on the characteristics corresponding to the third network element. The meaning of the characteristics corresponding to the third network element can be found in the preceding description of the characteristics corresponding to the third network element and is not listed here. Similarly, the at least one third network element can send fourth information to the first network element. Similarly, the first network element receives at least one fourth information in total. The fourth information received from one of the at least one third network element indicates the characteristics corresponding to each third network element.

[0243] Under D3, the method for determining the M third network elements can refer to the content of determining the M third network elements under D1 above, and the repeated parts are not listed again.

[0244] When the first network element determines M third network elements, it can optionally also determine the features of each of the M third network elements used to participate in the distributed learning model training. The features of each third network element used to participate in the distributed learning model training may be part or all of the features corresponding to each third network element.

[0245] In one possible design, the first network element may send the second information to M third network elements respectively. By analogy, the first network element sends a total of M second information. The first network element may send the second information directly to the M third network elements. Alternatively, the first network element may send the second information to the M third network elements through a collaborating party (such as a fourth network element). Alternatively, if the first network element acts as both an initiator and a collaborating party, the first network element may also send the second information directly to the M third network elements. The second information sent to one of the M third network elements is used to indicate the characteristics of the third network element for participating in distributed learning model training. Optionally, the characteristics for participating in distributed learning model training are used to indicate that the third network element participates in distributed learning model training. Alternatively, the second information may also indicate that the third network element participates in distributed learning model training.

[0246] Further, optionally, M third network elements may respectively send first response messages to the first network element. By analogy, the first network element receives a total of M first response messages. The first response message sent by one of the third network elements to the first network element is used to indicate that the third network element participates in the distributed learning model training (or is described as accepting participation in the distributed learning model training), or does not participate in the distributed learning model training (or is described as not accepting participation in the distributed learning model training). In the case where the second information received by a third network element also indicates a feature of a third network element for participating in distributed learning model training, the first response message sent by the third network element can also be specifically understood as accepting the feature for participating in the distributed learning model training for distributed learning model training, or not accepting participation in the feature for participating in the distributed learning model training for distributed learning model training.

[0247] In an embodiment of the present application, a mechanism for negotiating (or aligning) network elements participating in distributed learning model training is provided. A first network element can flexibly discover other participants in the distributed learning model training through a second network element. Furthermore, the first network element and the determined M third network elements can also negotiate or align features used to participate in the distributed learning model training to ensure the accuracy of subsequent distributed learning model training.

[0248] The communication method involved in FIG7 is described below by taking the first request indicating one or more characteristics as an example, in conjunction with the schematic diagram of a communication method shown in FIG8. The number of the at least one third network element involved in FIG8 can be arbitrary and is not limited thereto.

[0249] S801: A first network element sends fifth information to a second network element. Correspondingly, the second network element receives the fifth information from the first network element. The fifth information indicates features supported by the first network element.

[0250] The fifth information can be understood as being used to request registration of the first network element. The fifth information can also be referred to as a registration request for the first network element. The features supported by the first network element can refer to the features corresponding to the third network element described above and are not listed here one by one. For example, the fifth information indicates the names (Feature name(s)) and / or identifiers of the features supported by the first network element. In addition to indicating the features supported by the first network element, the fifth information optionally also indicates at least one of the identifier of the first network element, the network element type, the network element capability (capability), the network element type (NF Type(s)) corresponding to the first network element, or the network element set identifier (NF Set ID(s)) corresponding to the first network element. The identifier of the first network element can refer to the identifier of the third network element described above and is not listed here. The type of the first network element, for example, indicates that the first network element is an AF. The network element capability of the first network element, for example, indicates that the network element has at least one of distributed learning capability, distributed learning model training or inference capability, etc. The network element set identifier refers to the identifier of the network element that the first network element supports for acquiring data.

[0251] The following is an example of the content of the fifth information.

[0252] [Corrected 04.03.2025 according to Rule 91] G1. The fifth information includes the network element type of the first network element, the network element set identifier corresponding to the first network element, the network element type corresponding to the first network element, and the event identifier corresponding to the feature supported by the first network element.

[0253] For example, if the first network element is NWDAF and the network element type corresponding to the first network element includes AMF, then it means that NWDAF supports collecting data from AMF, or if the first network element is NWDAF and the network element type corresponding to the first network element includes SMF, then it means that NWDAF supports collecting data from SMF. Also, if the first network element is NWDAF and Event ID = location report, it means that NWDAF supports collecting UE location information from AMF; for another example, Event ID = Change of RAT Type, it means that NWDAF supports collecting RAT Type change information from AMF. For another example, if the network elements corresponding to the network element set identifier 1 corresponding to the first network element include AMF1, AMF2 and AMF3, then it means that the first network element supports collecting data from AMF1, AMF2 and AMF3.

[0254] [Corrected 04.03.2025 according to Rule 91] G2. The fifth information includes the feature name and feature identifier. The meaning of the feature identifier can be referred to in the previous discussion of the feature identifier in Figure 7, and the repeated parts are not listed here. The feature name may be, for example, service experience, location, average throughput, average packet delay, etc. For example, Feature ID = 1234 indicates that the feature is average throughput, and Feature ID = 1235 indicates that the feature is average packet delay.

[0255] [Corrected 04.03.2025 according to Rule 91] The above G1 or G2 can be applicable to the first network element, which can be any type of network element such as NWDAF, 5GC NF (such as AMF, SMF, UPF, PCF, etc.) or AF, without limitation.

[0256] After receiving the fifth information, the second network element may store the fifth information and send a first registration response to the first network element. The first registration response indicates that the first network element has successfully registered or that the first network element has not successfully registered.

[0257] If the first network element has been registered in the second network element before, or the first network element does not need to be registered in the second network element, step S801 does not need to be executed, that is, S801 is an optional step, which is indicated by a dotted line in Figure 8.

[0258] S802: At least one third network element sends sixth information to the second network element. Correspondingly, the second network element receives the sixth information from the at least one third network element.

[0259] One of the sixth information indicates the features supported by the third network element corresponding to the sixth information. The sixth information can be understood as being used to request registration of the third network element. For example, the sixth information indicates the name and / or identifier of the features supported by the third network element. In addition to indicating the features supported by the third network element corresponding to the sixth information, one of the sixth information optionally also indicates the identifier of the third network element. In addition to indicating the features supported by the third network element, the sixth information optionally also indicates at least one of the identifier of the third network element, the network element type, the capability, the network element type corresponding to the third network element, or the network element set identifier corresponding to the third network element. The capabilities of the third network element, the network element type corresponding to the third network element, and the network element set identifier corresponding to the third network element can refer to the capabilities of the first network element, the network element type corresponding to the first network element, and the network element set identifier corresponding to the first network element discussed above, respectively.

[0260] In cases where at least one third network element has been previously registered in the second network element, step S802 may not be performed, that is, S802 is an optional step, which is indicated by a dotted line in FIG8 .

[0261] The execution order of S801 and S802 may be arbitrary, for example, they may be executed simultaneously, or S801 may be executed first and then S802, or S802 may be executed first and then S802, and there is no specific limitation on this.

[0262] S803. The first network element sends a first request to the second network element. Correspondingly, the second network element receives the first request from the first network element. In the embodiment of the present application, the first request is used to request network element discovery and indicates one or more characteristics. Of course, the first request may also indicate other information (such as one or more network element types), which is not limited to this.

[0263] Among them, the content of the first request and the content of one or more features can refer to the content of the first request and the content of one or more features discussed in Figure 7 above, and the repeated parts are not listed again.

[0264] S804: The second network element sends first information to the first network element. Accordingly, the first network element receives the first information from the second network element. The first information includes information about at least one network element group. The information about the at least one network element group is used to indicate at least one third network element.

[0265] The content of the first information, the information of at least one network element group, and the content of at least one third network element can refer to the content of the first information, the information of at least one network element group, and the content of at least one third network element discussed in Figure 7 above, and the repeated parts are not listed again.

[0266] S805. The first network element determines M third network elements from at least one network element group according to the first information.

[0267] The contents of the M third network elements and the contents for determining the M third network elements may refer to the contents of the M third network elements and the contents for determining the M third network elements discussed in FIG. 7 above, respectively, and the repetitive parts will not be listed again.

[0268] In another possible implementation, the first network element may also directly use at least one third network element as a participant in the distributed learning model training. In this case, the first network element does not need to execute step S805, that is, S805 is an optional step, which is indicated by a dotted line in Figure 8.

[0269] S806: The first network element sends second information to the M third network elements. Correspondingly, the M third network elements receive the second information from the first network element. The second information indicates features for participating in distributed learning model training.

[0270] The first network element may directly send the second information to the M third network elements respectively, or the first network element may send the second information to the M third network elements respectively through a collaborative party (such as the fourth network element), and there is no specific limitation on this.

[0271] The content of the second information, the content of the features used to participate in the distributed learning model training, and the content of sending the second information can be referred to the content of the second information, the content of the features used to participate in the distributed learning model training, and the content of sending the second information discussed in Figure 7 above, and the repetitions will not be listed again.

[0272] In one possible implementation, the first network element determines to trigger a distributed learning model training preparation phase and determines second information. The second information can be regarded as preparation information for distributed learning model training. For example, if the distributed learning model training is VFL model training, the preparation information can be referred to as the VFL Training Preparation Information.

[0273] In one possible design, the content of a second message sent to a third network element indicating the features of the third network element for participating in the distributed learning model training can refer to the content of a second message indicating the features of the third network element for participating in the distributed learning model training discussed in FIG7 above, and the repeated parts are not listed again. Optionally, a second message sent to a third network element may include at least one of the following information: a type identifier of the distributed learning model training, a preparation stage identifier, a distributed learning model training ID, model requirements, or alignment information (alignment information) of the distributed learning model training of each third network element, etc. The alignment information includes samples and / or features of the distributed learning model training of each third network element, etc. When the second information includes alignment information, and the alignment information includes the features of the distributed learning model training of each third network element, it is equivalent to the second information indicating the features of the distributed learning model training of each third network element.

[0274] The distributed learning model training type identifier indicates the type of distributed learning model training, such as longitudinal federated learning between NWDAFs and AFs, longitudinal federated learning between NWDAFs, or longitudinal federated learning between AFs assisted by NWDAFs. The preparation phase identifier indicates the preparation phase of the distributed learning model training to be performed. The distributed learning model training ID is a unique identifier for the training process. Model requirements include, for example, the output category of the model, such as discrete values, such as those associated with classification algorithms, or actual values, such as those associated with regression algorithms.

[0275] In this way, the initiator exchanges information with each participant to coordinate the model training process to be performed by each participant. During the training preparation phase, all network elements receive preparation information, which indicates that the network element (e.g., passive participant) should link its local model training with the initiator's model training. This enables each participant to map and control the information required to implement distributed learning model training.

[0276] As an embodiment, S806 can be used as an independent embodiment, and the contents of FIG. 8 except S806 can be used as optional implementation methods of this embodiment.

[0277] S807: The M third network elements respectively send first response messages to the first network element. Correspondingly, the first network element respectively receives first response messages from the M third network elements. The first response messages indicate whether to participate in or not to participate in the distributed learning model training.

[0278] Each of the M third network elements determines whether the distributed learning model training process requested by the first network element can be executed based on the received second information and the relevant configurations available locally. Relevant configurations include at least one of the types of distributed learning model training that can be participated in (for example, in NWDAF; NWDAF-AF; or NWDAF Assisting AFs, etc.), the roles of supported distributed learning model training (such as master participant, slave participant or collaborator, etc.), or the scope information of the distributed learning model. The type of distributed learning model training can be understood as the type of participant supported according to the distributed learning model training process. The scope information of the distributed learning model, for example, the interoperability indicators of the distributed learning model training of the suppliers participating in the task, the supported analysis filter information (such as area of ​​interest (AoI), slice identification or application identification).

[0279] If a third network element can execute the distributed learning model training process requested by the first network element, then the first network element can perform the requested preparatory work, and the third network element can send a first response message to the first network element indicating participation. If the distributed learning model training process requested by the first network element cannot be executed, then the third network element can send a first response message to the first network element indicating that it does not participate in the distributed learning model training. Of course, the third network element can also send the first response message directly to the first network element, or it can send the first response message through a collaborating party (such as a fourth network element). The rest of the content of the first response message can refer to the content of the first response message discussed in Figure 7 above, and the repetitive parts will not be listed again.

[0280] After S807, the first network element determines the final participants in the distributed learning model training based on the first response messages from the M third network elements. For example, if a third network element indicates in the first response message that it does not participate in the distributed learning model training, the first network element will not select that third network element for participation in the distributed learning model training. If a third network element indicates in the first response message that it participates in the distributed learning model training, the first network element may select that third network element for participation in the distributed learning model training. This process is repeated and so on, thereby determining the participants in the distributed learning model training.

[0281] The embodiments of the present application provide a mechanism whereby a first network element can indicate one or more desired features to a second network element to discover participants that support the corresponding features, thereby completing the feature alignment process. This improves the flexibility of determining participants and the flexibility of feature alignment. Furthermore, the embodiments of the present application also provide a mechanism for participants (e.g., the first network element and at least one third network element) to register with the second network element.

[0282] The communication method involved in FIG7 is described below by taking the first request indicating one or more characteristics as an example, in conjunction with the schematic diagram of a communication method shown in FIG9. The number of the at least one third network element involved in FIG9 can be arbitrary and is not limited thereto.

[0283] S901: A first network element sends fifth information to a second network element. Correspondingly, the second network element receives the fifth information from the first network element. The fifth information indicates features supported by the first network element.

[0284] The content of the fifth information can refer to the content of the fifth information discussed in Figure 8 above, and the repeated parts are not listed again.

[0285] After receiving the fifth information, the second network element may store the fifth information and send a first registration response to the first network element. The first registration response indicates that the first network element has successfully registered or that the first network element has not successfully registered.

[0286] If the first network element has been registered in the second network element before, or the first network element does not need to be registered in the second network element, step S901 does not need to be executed, that is, S901 is an optional step, which is indicated by a dotted line in Figure 9.

[0287] S902: At least one third network element sends sixth information to the second network element. Correspondingly, the second network element receives the sixth information from the at least one third network element.

[0288] The content of the sixth information can refer to the content of the sixth information discussed in Figure 8 above, and the repeated parts are not listed again.

[0289] In cases where at least one third network element has been previously registered in the second network element, step S902 may not be required, that is, S902 is an optional step, which is indicated by a dotted line in FIG9 .

[0290] The execution order of S901 and S902 may be arbitrary, for example, they may be executed simultaneously, or S901 may be executed first and then S902, or S902 may be executed first and then S902, and there is no specific limitation on this.

[0291] S903. The first network element sends a first request to the second network element. Correspondingly, the second network element receives the first request from the first network element. In the embodiment of the present application, the first request is used to request the discovery of network elements, and indicates one or more network element types, and the characteristics corresponding to each network element type. The characteristics corresponding to one or more network element types are one or more characteristics. S903 can be understood as the first network element specifying the characteristics of the network element type in the first request, or can be described as specifying the network element type corresponding to the desired characteristics.

[0292] Among them, the content of the first request, the content of one or more network element types, and the content of one or more characteristics can refer to the content of the first request, the content of one or more network element types, and the content of one or more characteristics discussed in Figure 7 above, and the repeated parts will not be listed again.

[0293] S904: The second network element sends the first information to the first network element. Accordingly, the first network element receives the first information from the second network element. The first information indicates at least one third network element. The content of the first information and the content of the at least one third network element can be referred to as the content of the first information and the content of the at least one third network element discussed in FIG. 7 , respectively. Any repetitions are omitted.

[0294] Exemplarily, because the first request specifies the network element type corresponding to the desired feature, the second network element can also discover network elements if the requirements of the corresponding network element type are met. In this case, the content of the first information can refer to the content of the first information discussed in Figure 7 above, and the repeated parts are not listed again. In the case where the first information includes information about at least one network element group, the network element types of the network elements included in one network element group all belong to one or more network element types. Optionally, a network element group may include features corresponding to network elements under one or more network element types.

[0295] For example, take the case where one or more network element types include NWDAF, and the first information includes information on 3 network element groups (such as network element group 1, network element group 2, and network element group 3). The information of network element group 1 includes {NWDAF1: Event ID 1, Event ID 4>, <NWDAF2: Event ID 2, Event ID 5}, the information of network element group 2 includes {NWDAF3: Event ID 1, Event ID 2, Event ID 4, Event ID 5}, and the information of network element group 3 includes {NWDAF4: Event ID 1, Event ID 2, Event ID 4, Event ID 5}.

[0296] Alternatively, take the case where one or more network element types include AF, and the first information includes information on 3 network element groups (such as network element group 1, network element group 2, and network element group 3). The information of network element group 1 includes {AF1: Feature ID3, AF2: Feature ID4}, the information of network element group 2 includes {AF3: Feature ID3, Feature ID4}, and the information of network element group 3 includes {AF4: Feature ID3, Feature ID4}.

[0297] S905. The first network element determines M third network elements from at least one third network element according to the first information.

[0298] The content of the M third network elements and the content of determining the M third network elements can be respectively referred to the content of the M third network elements and the content of determining the M third network elements discussed in Figure 7 above. Duplicate parts will not be listed again.

[0299] In another possible implementation, the first network element can also directly use at least one third network element as a participant in the distributed learning model training. In this case, the first network element does not need to perform the steps of S905, that is, S905 is an optional step, which is shown by a dotted line in Figure 9.

[0300] S906. The first network element sends the second information to the M third network elements. Correspondingly, the M third network elements receive the second information from the first network element. The second information indicates the features used to participate in the distributed learning model training.

[0301] The content of the second information, the content of the features used to participate in the distributed learning model training, and the content of sending the second information can be respectively referred to the content of the second information and the content of the features used to participate in the distributed learning model training discussed in Figure 8 above. Duplicate parts will not be listed again.

[0302] S907. The M third network elements each send a first response message to the first network element. Accordingly, the first network element receives the first response messages from the M third network elements. The first response message indicates whether to participate in the distributed learning model training. The content of the first response message can be referred to the content of the first response message discussed in FIG. 7 above, and any repetition is not listed here.

[0303] After S907, the first network element determines the final participants in the distributed learning model training based on the first response messages from the M third network elements. For example, if a third network element indicates in the first response message that it does not participate in the distributed learning model training, the first network element will not select that third network element for participation in the distributed learning model training. If a third network element indicates in the first response message that it participates in the distributed learning model training, the first network element may select that third network element for participation in the distributed learning model training. This process is repeated and so on, thereby determining the participants in the distributed learning model training.

[0304] In the embodiments of the present application, a first network element can indicate one or more desired features and one or more network element types to a second network element to discover participants that support the corresponding features and conform to the corresponding network element types. This improves the flexibility of determining participants and enables the determined participants to meet the requirements of distributed learning model training. Furthermore, the first network element can also align features with M third network elements, providing a feature alignment mechanism.

[0305] To improve the flexibility of determining participants, an embodiment of the present application also provides a communication method. In this method, the initiator of a distributed learning (such as a first network element) can determine at least one third network element, and then the first network element can request supported features from the at least one third network element, and based on the features supported by the at least one third network element, determine the participants (such as M network elements) who will ultimately participate in the distributed learning from the at least one third network element. In this way, there is no need to predetermine the participants, and a mechanism for online determination of distributed participants is implemented, which helps to improve the flexibility of determining participants.

[0306] The communication method provided in the embodiments of the present application is introduced below with reference to the accompanying drawings.

[0307] FIG10 illustrates a communication method provided in an embodiment of the present application. The following describes the various steps involved in FIG10 .

[0308] S1001. A first network element determines at least one third network element.

[0309] At least one third network element can be considered as a candidate network element for participating in the training of the distributed learning model. The first network element may pre-store information of at least one third network element, such as the identifier or address of at least one third network element. The identifier of one third network element can refer to the identifier of the third network element discussed in Figure 7 above, and the repeated parts are not listed again. In this way, the first network element can determine the at least one third network element based on the pre-stored information of the at least one third network element. Alternatively, the first network element can receive fourth information from the second network element, and the fourth information indicates information of the at least one third network element. In this way, the first network element can determine the at least one third network element based on the fourth information.

[0310] S1002: The first network element sends first information to at least one third network element, and correspondingly, the at least one third network element receives the first information from the first network element.

[0311] The first information sent to a third network element is used to request the third network element to support features, or can be described as being used to request the third network element to provide feedback on the features it supports. The details of the features supported by the third network element can be found in the discussion of the features supported by the third network element in FIG. 7 , and any repetitions are omitted. Optionally, the first information includes a feature reporting indication, which is used to request the third network element to provide feedback on the features it supports.

[0312] S1003: At least one third network element sends second information to the first network element. Correspondingly, the first network element receives the second information from the at least one third network element. The second information sent by one of the third network elements to the first network element indicates features supported by the third network element.

[0313] For example, the manner in which one of the second information indicates the features supported by the third network element corresponding to the second information can refer to the content of the first request indicating one or more features discussed above, and will not be listed one by one here. For example, a third network element can register coarse-grained features (such as NF type(s) and / or NF set ID(s), or event ID(s)) with the first network element, or a third network element can register fine-grained features with the first network element.

[0314] S1004. The first network element determines features of M third network elements for participating in distributed learning model training.

[0315] The manner in which the first network element determines the M third network elements can refer to the content of determining the M third network elements discussed in FIG7 , and the repeated parts are not listed again. Alternatively, the first network element can use at least one third network element as the M third network elements.

[0316] Since the second information indicates the features supported by at least one third network element, the first network element can determine the features of M third network elements for participating in the distributed learning model training according to the requirements of the distributed learning model training.

[0317] For example, when the first network element determines at least one third network element, it is already able to clearly identify the fine-grained features supported by the third network element. Then, the first network element can determine the features of M third network elements used to participate in the distributed learning model training from the fine-grained features supported by the M third network elements. Alternatively, if the second information indicates that the third network element needs to support coarse-grained features, then the first network element can negotiate with the third network element through S1002 and S1003 on the fine-grained features (such as feature(s) or feature ID(s) or event ID(s)) needed for model training, and then determine the features of M third network elements used to participate in the distributed learning model training.

[0318] In one possible implementation, the M third network elements each send a first response message to the first network element. Correspondingly, the first network element receives first response information from the M third network elements. The first response message indicates whether to participate in or not to participate in the distributed learning model training.

[0319] Each of the M third network elements determines whether it can execute the distributed learning model training process requested by the first network element based on the received second information and locally available relevant configurations. The relevant configurations may include at least one of the types of distributed learning model training that can be participated in, supported VFL roles, or scope information of the distributed learning model.

[0320] If a third network element can execute the distributed learning model training process requested by the first network element, then the first network element can perform the requested preparatory work, and the third network element can send a first response message to the first network element indicating participation. If the distributed learning model training process requested by the first network element cannot be executed, then the third network element can send a first response message to the first network element indicating that it does not participate in the distributed learning model training. Of course, the third network element can also send the first response message directly to the first network element, or it can send the first response message through a collaborative party (such as a fourth network element).

[0321] In the embodiments of the present application, the first network element can directly determine at least one third network element and determine the participants who will ultimately participate in the distributed learning model training from these third network elements, thereby increasing the flexibility of determining the participants. In addition, at least one third network element can be registered with the second network element, but the features supported by the at least one third network element are not registered, thereby reducing the risk of leakage of the features supported by the at least one third network element.

[0322] Taking the example of the first network element determining at least one third network element based on the fourth information, the communication method involved in Figure 10 is introduced by way of example in combination with the schematic diagram of a communication method shown in Figure 11.

[0323] S1101: A first network element sends fifth information to a third network element. The fifth information indicates information of the first network element. Correspondingly, the third network element receives the fifth information from the first network element. The fifth information indicates information of the first network element.

[0324] The information of the first network element includes the identifier or address of the first network element. Optionally, the information of the first network element also includes the network element type and / or network element capability of the first network element. The content of the network element capability can refer to the content of the network element capability discussed in Figure 8 above, and the repeated parts are not listed again.

[0325] S1102: At least one third network element sends sixth information to the second network element. Correspondingly, the second network element receives the sixth information from the at least one third network element. The sixth information sent by one of the third network elements to the second network element indicates information of the third network element.

[0326] The information of the third network element includes an identifier or address of the third network element. Optionally, the information of the third network element also includes a network element type and / or network element capability of the third network element.

[0327] S1103: The second network element sends fourth information to the first network element. Accordingly, the first network element receives the fourth information from the second network element. The fourth information indicates information about at least one third network element. The information about the at least one third network element includes an identifier or address of the at least one third network element.

[0328] S1104. The first network element determines at least one third network element according to the fourth information.

[0329] Since the fourth information indicates information of at least one third network element, the first network element can directly determine the at least one third network element according to the fourth information.

[0330] S1105. The first network element sends first information to at least one third network element. Accordingly, the at least one third network element receives the first information from the first network element. The first information sent to a third network element is used to request features supported by the third network element, or can be described as requesting the third network element to send (or provide feedback on) features supported by the third network element, or can be described as requesting to obtain features supported by the third network element.

[0331] S1106: At least one third network element sends second information to the first network element. Correspondingly, the first network element receives the second information from the at least one third network element. The second information indicates features supported by the third network element.

[0332] For example, the second information includes an identifier or an address of the third network element, and features supported by the third network element.

[0333] S1107. The first network element determines, based on the second information, features of M third network elements for participating in distributed learning model training.

[0334] The content of the features determined by the first network element for M third network elements to participate in the distributed learning model training can refer to the content of the features determined by the first network element for M third network elements to participate in the distributed learning model training discussed in Figure 10 above, and the repetitions will not be listed again.

[0335] For example, at least one third network element may register coarse-grained features (e.g., NF type(s) and / or NF set ID(s), or event ID(s)) with the first network element. The first network element may discover at least one third network element through the second network element. The network element discovery request indicates the coarse-grained features that the third network element needs to support. After discovering the third network element, the first network element negotiates with the third network element through S1105 and S1106 on the fine-grained features (e.g., feature(s) or feature ID(s) or event ID(s)) required for model training. Furthermore, the features of M third network elements used to participate in distributed learning model training are determined.

[0336] S1108. The first network element sends third information to M third network elements of the at least one third network element. Accordingly, the M third network elements receive the third information from the first network element. The third information sent to one of the third network elements indicates features used by the third network element to participate in distributed learning model training. Feature alignment is achieved in this manner.

[0337] The first network element may directly send the third information to each of the M third network elements, or the first network element may send the third information to each of the M third network elements through a collaborating party (such as a fourth network element), without specific limitation. For example, the first network element may determine a feature of each of the at least one third network element for participating in the distributed learning model training, and indicate the feature of the third network element for participating in the distributed learning model training in the third information sent by one of the third network elements.

[0338] In one possible implementation, the first network element determines to trigger the distributed learning model training preparation phase and determines third information. The third information can be regarded as preparation information for the distributed learning model training. For example, if the distributed learning model training is VFL model training, the preparation information can be referred to as the VFL Training Preparation Information.

[0339] In one possible design, the manner in which a third information sent to a third network element indicates the characteristics of the third network element for participating in distributed learning model training can refer to the content of the second information indicating the characteristics of the third network element for participating in distributed learning model training discussed in FIG. 7 above, and is not listed here. Optionally, the third information may include at least one of the following information: a type identifier of distributed learning model training, a preparation stage identifier, a distributed learning model training ID, model requirements, or alignment information (alignment information) of distributed learning model training for each third network element. The type identifier of distributed learning model training, the preparation stage identifier, the distributed learning model training ID, the model requirements, and the content of the alignment information of distributed learning model training for each third network element can refer to the type identifier of distributed learning model training, the preparation stage identifier, the distributed learning model training ID, the model requirements, and the content of the alignment information of distributed learning model training for each third network element discussed in FIG. 8 above, respectively, and are not listed here. When the third information includes alignment information, and the alignment information includes the characteristics of distributed learning model training for each third network element, it is equivalent to the third information indicating the characteristics of distributed learning model training for each third network element.

[0340] As an embodiment, S1108 can be used as an independent embodiment, and the contents of FIG11 except S1108 can be used as optional implementation methods of this embodiment.

[0341] S1109. The M third network elements send a first response message to the first network element. Accordingly, the first network element receives the first response message sent by at least one third network element. The first response message is used to indicate whether to participate in the distributed learning model training. The content of the first response message can be referred to the content of the first response message discussed in FIG. 10 above, and any repetition is not listed here.

[0342] After S1109, the first network element determines the final participants in the distributed learning model training based on the first response messages from the M third network elements. For example, if a third network element indicates in the first response message that it does not participate in the distributed learning model training, the first network element will not select that third network element for participation in the distributed learning model training. If a third network element indicates in the first response message that it participates in the distributed learning model training, the first network element may select that third network element for participation in the distributed learning model training. This process is repeated and so on, thereby determining the participants in the distributed learning model training.

[0343] In this embodiment of the present application, participants do not need to register their supported features with the second network element, reducing the risk of feature leakage. A first network element can discover at least one third network element through the second network element and request the at least one third network element to report its supported features. The first network element then determines M third network elements to participate in vertical federated learning based on the features reported by the at least one third network element. This increases the flexibility of determining the features that participate in distributed learning model training. Furthermore, the first network element can determine the features that the M third network elements use to participate in the distributed learning model, thereby completing the feature alignment process.

[0344] It is understood that, in order to implement the functions in the above embodiments, each network element includes a hardware structure and / or software module corresponding to the execution of each function. Those skilled in the art should readily appreciate that, in conjunction with the various exemplary units and method steps described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in hardware or in a manner driven by computer software depends on the specific application scenario and design constraints of the technical solution.

[0345] Please refer to Figure 12, which is a structural diagram of a communication device provided in an embodiment of the present application. The communication device can be used to implement the functions of the first network element, the second network element or any third network element in the above method embodiment, and thus can also achieve the beneficial effects possessed by the above method embodiment. In the embodiment of the present application, the communication device can be the master participant or slave participant involved in Figure 2, the initiator or the first network element involved in Figure 4, the initiator or AF involved in Figure 6, the registrant or the second network element involved in Figure 4, any network element involved in Figure 5 (such as VFLSF), the registrant or VFLSF involved in Figure 6, the master participant or slave participant involved in Figure 2, the registrant or the second network element involved in Figure 4, any network element involved in Figure 5, or other participants or any third network element involved in Figure 6, etc.

[0346] As shown in Figure 12, the communication device 1200 includes a processing module 1210 and a transceiver module 1220. The communication device 1200 is used to implement the function of the first network element in any of the method embodiments shown in Figures 7 to 9 above, implement the function of the second network element in any of the method embodiments shown in Figures 7 to 9 above, implement the function of the first network element in the method embodiment shown in Figures 10 or 11 above, or implement the function of any third network element in the method embodiment shown in Figures 10 or 11 above.

[0347] In the first embodiment, the communication device 1200 is used to implement the function of the first network element in any one of the method embodiments shown in FIG. 7 to FIG. 9 .

[0348] For example, the transceiver module 1220 is configured to send a first request and receive first information under the control of the processing module 1210. Optionally, the processing module 1210 is further configured to determine M third network elements.

[0349] In the second embodiment, the communication device 1200 is used to implement the function of the second network element in any one of the method embodiments shown in FIG. 7 to FIG. 9 .

[0350] For example, the transceiver module 1220 is configured to receive a first request and send first information under the control of the processing module 1210 .

[0351] In the third embodiment, the communication device 1200 is used to implement the function of the first network element in the method embodiment shown in FIG. 10 or FIG. 11 .

[0352] For example, the transceiver module 1220 is used to send the first information and receive the second information; the processing module 1210 is used to determine at least one third network element and determine the characteristics of M third network elements for participating in the distributed learning model training.

[0353] In a fourth embodiment, the communication device 1200 is used to implement the function of the third network element in the method embodiment shown in FIG. 10 or FIG. 11 .

[0354] For example, the transceiver module 1220 is configured to receive first information and send second information under the control of the processing module 1210 .

[0355] A more detailed description of the processing module 1210 and the transceiver module 1220 can be directly obtained by referring to the relevant description in any of the method embodiments shown in Figures 7 to 11, and will not be repeated here.

[0356] Please refer to Figure 13, which is a structural diagram of a communication device provided in an embodiment of the present application. The communication device can be used to implement the functions of the first network element, the second network element, or any third network element in the above-mentioned method embodiment, and thus can also achieve the beneficial effects possessed by the above-mentioned method embodiment. In the embodiment of the present application, the communication device can be the master participant or slave participant involved in Figure 2, the initiator or the first network element involved in Figure 4, the initiator or AF involved in Figure 6, the registrant or the second network element involved in Figure 4, any network element involved in Figure 5 (such as VFLSF), the registrant or VFLSF involved in Figure 6, the master participant or slave participant involved in Figure 2, the registrant or the second network element involved in Figure 4, any network element involved in Figure 5, or other participants or any third network element involved in Figure 6, etc. As shown in Figure 13, the communication device 1300 includes a processing circuit 1310 and an interface circuit 1320. The processing circuit 1310 and the interface circuit 1320 are coupled to each other. The implementation of processing circuit 1310 can refer to the content discussed above, and the repetitions are not listed here. Interface circuit 1320 can be a transceiver or an input / output interface. Optionally, communication device 1300 can also include memory 1330 for storing instructions executed by processing circuit 1310, storing input data required by processing circuit 1310 to execute instructions, or storing data generated after processing circuit 1310 executes instructions.

[0357] [Corrected 04.03.2025 according to Rule 91] Communication device 1300 may be used to implement any of the method embodiments shown in Figures 7 to 11. Optionally, communication device 1300 may be used to implement the functions of communication device 1200 shown in Figure 12 above. For example, processing circuit 1310 may be used to implement the functions of processing module 1210, and interface circuit 1320 may be used to implement the functions of transceiver module 1220.

[0358] When the communication device is a chip applied to a network element (such as the first network element, the second network element, or the third network element), the network element chip implements the functions of the network element in the above method embodiment. The network element chip receives information from another module (such as a radio frequency module or an antenna) in the network element, where the information is sent from the other network element to the network element; or the network element chip sends information to another module (such as a radio frequency module or an antenna) in the network element, where the information is sent from the network element to the other network element.

[0359] The memory involved in various embodiments of the present application may include volatile memory, such as random access memory (RAM). The memory may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0360] An embodiment of the present application provides another example of a communication device, which includes at least one processor and at least one memory, the at least one processor and the at least one memory being coupled, the at least one memory being used to store instructions, and when the instructions are executed by the at least one processor, the communication device executes the method in the above embodiment. Taking the communication device including a processor and a memory as an example, as shown in Figure 14, the communication device 1400 includes a processor 1410 and a memory 1420. The processor 1410 and the memory 1420 are coupled, and the memory 1420 stores instructions. When the instructions stored in the memory 1420 are executed by the processor 1410, the communication device 1400 executes the method in any of the above embodiments.

[0361] The processor involved in each embodiment of the present application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor can be a microprocessor or any conventional processor.

[0362] The method steps in each embodiment of the present application can be implemented in hardware or in software instructions that can be executed by a processor. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disk, mobile hard disk, CD-ROM or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. The storage medium can also be an integral part of the processor. The processor and storage medium can be located in an ASIC. In addition, the ASIC can be located in a base station or a terminal. The processor and storage medium can also exist in a base station or a terminal as discrete components.

[0363] In the above embodiments, all or part of the embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are performed in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, or other programmable device. The computer program or instructions may be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions may be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; an optical medium, such as a digital video disk; or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or nonvolatile storage medium, or may include both volatile and nonvolatile types of storage media.

[0364] An embodiment of the present application provides a communication system, comprising a first network element and a second network element, wherein the first network element can implement the functions of the first network element in any of the method embodiments described in Figures 7 to 9 above, and the second network element can implement the functions of the second network element in any of the method embodiments described in Figures 7 to 9 above. Optionally, the communication device further comprises at least one third network element, wherein the at least one third network element can implement the functions of the at least one third network element in any of the method embodiments described in Figures 7 to 9 above.

[0365] An embodiment of the present application provides a communication system, comprising a first network element and at least one third network element, wherein the first network element can implement the functions of the first network element in the method embodiment in FIG. 10 or FIG. 11 , and the at least one third network element can implement the functions of the at least one third network element in the method embodiment in FIG. 10 or FIG. 11 Optionally, the communication system further comprises a second network element, which can implement the functions of the second network element in the method embodiment in FIG. 10 or FIG. 11 .

[0366] An embodiment of the present application provides a chip system, comprising: a processor and an interface, wherein the processor is configured to call and execute instructions from the interface, and when the processor executes the instructions, the method described in any one of FIG. 7 to FIG. 11 is implemented.

[0367] An embodiment of the present application provides a computer-readable storage medium for storing computer programs or instructions, which, when executed, implements the method described in any one of Figures 7 to 11 above.

[0368] An embodiment of the present application provides a computer program product comprising instructions, which, when executed on a computer, implements any one of the methods described in FIG. 7 to FIG. 11 .

[0369] In the various embodiments of the present application, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced by each other. The technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationships.

[0370] It should be understood that the various numbers used in the various embodiments of this application are merely for ease of description and are not intended to limit the scope of the embodiments of this application. The order of the sequence numbers of the above-mentioned processes does not necessarily imply a specific order of execution; the order of execution of the processes should be determined by their functions and inherent logic.

Claims

1. A communication method, characterized in that: Applied to vertical federated learning servers, including: Sending a first request to a network storage function network element, where the first request is used to request discovery of a vertical federated learning client, and the first request indicates one or more feature identifiers, where the one or more feature identifiers indicate features of the vertical federated learning client requested to be discovered for vertical federated learning model training; Receive first information from the network storage function network element, where the first information includes information of at least one vertical federated learning client, wherein a union of feature identifiers supported by the at least one vertical federated learning client includes the one or more feature identifiers.

2. The method according to claim 1, characterized in that The first request further indicates one or more network element types, and a feature identifier corresponding to each of the one or more network element types, where the feature identifier corresponding to each network element type belongs to the one or more feature identifiers; The network element type of the at least one vertical federated learning client belongs to the one or more network element types.

3. The method according to claim 1 or 2, characterized in that The first information further indicates a feature identifier corresponding to each vertical federated learning client in the at least one vertical federated learning client.

4. The method according to claim 3, characterized in that The feature identifier corresponding to each vertical federated learning client in the at least one vertical federated learning client is a feature identifier in the one or more feature identifiers.

5. The method according to claim 3 or 4, characterized in that The method further comprises: Based on the at least one vertical federated learning client, M vertical federated learning clients are determined, and the M vertical federated learning clients are used to participate in vertical federated learning model training.

6. The method according to claim 5, characterized in that The method further comprises: Determine a feature identifier for each of the M vertical federated learning clients to participate in the vertical federated learning model training; Second information is sent to each of the M vertical federated learning clients, wherein the second information sent to one of the M vertical federated learning clients is used to indicate a feature identifier of the one vertical federated learning client for participating in the training of the vertical federated learning model.

7. The method according to any one of claims 1 to 6, characterized in that The first information includes: Information about at least one vertical federated learning client group, wherein one vertical federated learning client group includes some or all of the at least one vertical federated learning client, and the union of feature identifiers supported by the vertical federated learning clients in the one vertical federated learning client group includes the one or more feature identifiers.

8. The method according to claim 7, characterized in that The first request further indicates one or more network element types, and a feature identifier corresponding to each of the one or more network element types, where the feature identifier corresponding to each network element type belongs to the one or more feature identifiers; Among them, the network element type of any vertical federated learning client in the vertical federated learning client group is the first network element type, the first network element type is one of the one or more network element types, and the union of feature identifiers supported by the vertical federated learning clients in the vertical federated learning client group includes the feature identifier corresponding to the first network element type indicated by the first request.

9. The method according to claim 7, characterized in that The first request further indicates one or more network element types, and a feature identifier corresponding to each of the one or more network element types, where the feature identifier corresponding to each network element type belongs to the one or more feature identifiers; Among them, the union of network element types of the vertical federated learning clients in the vertical federated learning client group includes the one or more network element types, the network element type of any vertical federated learning client in the vertical federated learning client group is one of the one or more network element types, and the union of feature identifiers supported by all vertical federated learning clients in the vertical federated learning client group whose network element type is the first network element type includes the feature identifier corresponding to the first network element type indicated by the first request, and the first network element type is one of the one or more network element types.

10. The method according to any one of claims 1 to 9, characterized in that The method further comprises: Sending third information to each of the at least one vertical federated learning clients, wherein the third information sent to one of the at least one vertical federated learning clients instructs the one vertical federated learning client to provide a feature identifier supported by the one vertical federated learning client; Fourth information is received respectively from the at least one vertical federated learning client, wherein the fourth information received from one of the at least one vertical federated learning client indicates a feature identifier supported by the one vertical federated learning client.

11. The method according to any one of claims 1 to 10, characterized in that The method further comprises: Send fifth information to the network storage function network element, where the fifth information indicates feature identifiers supported by the vertical federated learning server.

12. A communication method, characterized in that: Applied to a network storage function network element, the method includes: Receiving a first request from a vertical federated learning server, the first request being used to request discovery of a vertical federated learning client, the first request indicating one or more feature identifiers, the one or more feature identifiers indicating features of the vertical federated learning client requested for discovery used for vertical federated learning model training; First information is sent to the vertical federated learning server, where the first information includes information of at least one vertical federated learning client, wherein a union of feature identifiers supported by the at least one vertical federated learning client includes the one or more feature identifiers.

13. The method according to claim 12, characterized in that The first request further indicates one or more network element types, and a feature identifier corresponding to each of the one or more network element types, where the feature identifier corresponding to each network element type belongs to the one or more feature identifiers; The network element type of the at least one vertical federated learning client belongs to the one or more network element types.

14. The method according to claim 12 or 13, characterized in that The first information further indicates a feature identifier corresponding to each vertical federated learning client in the at least one vertical federated learning client.

15. The method according to claim 14, characterized in that The feature identifier corresponding to each vertical federated learning client in the at least one vertical federated learning client is a feature identifier in the one or more feature identifiers.

16. The method according to any one of claims 12 to 15, characterized in that: The first information includes: Information of at least one vertical federated learning client group, wherein the information of one vertical federated learning client group includes some or all of the vertical federated learning clients in the one vertical federated learning client group, and the union of feature identifiers supported by the vertical federated learning clients in the one vertical federated learning client group includes the one or more feature identifiers.

17. The method according to claim 16, characterized in that The first request further indicates one or more network element types, and a feature identifier corresponding to each of the one or more network element types, where the feature identifier corresponding to each network element type belongs to the one or more feature identifiers; Among them, the network element type of any vertical federated learning client in the vertical federated learning client group is the first network element type, the first network element type is one of the one or more network element types, and the union of feature identifiers supported by the vertical federated learning clients in the vertical federated learning client group includes the feature identifier corresponding to the first network element type indicated by the first request.

18. The method according to claim 16, characterized in that The first request further indicates one or more network element types, and a feature identifier corresponding to each of the one or more network element types, where the feature identifier corresponding to each network element type belongs to the one or more feature identifiers; The union of network element types of the vertical federated learning clients in the vertical federated learning client group includes the corresponding one or more network element types, the network element type of any vertical federated learning client in the vertical federated learning client group is one of the one or more network element types, and the union of feature identifiers supported by all vertical federated learning clients in the vertical federated learning client group whose network element types are all the first network element type includes the feature identifier corresponding to the first network element type indicated by the first request, and the first network element type is one of the one or more network element types.

19. The method according to any one of claims 12 to 18, characterized in that: The method further comprises: receiving fifth information from the vertical federated learning server, the fifth information indicating feature identifiers supported by the vertical federated learning server; and / or, Receive sixth information from the at least one vertical federated learning client, and receive sixth information from one of the at least one vertical federated learning client to indicate a feature identifier supported by the one vertical federated learning client.

20. A communication method, characterized in that: Applied to a first network element, the method includes: determining at least one third network element; Sending first information to each of the at least one third network elements, wherein the first information sent to one of the at least one third network element is used to request features supported by the one third network element, wherein some or all of the features supported by the at least one third network element are features corresponding to data trained by the distributed learning model; respectively receiving second information from the at least one third network element, wherein the second information received from one of the at least one third network element indicates a feature supported by the one third network element; Based on the second information received from the at least one third network element, determine the characteristics of M third network elements used to participate in the distributed learning model training, where the M third network elements are part or all of the at least one third network element, and M is a positive integer.

21. The method according to claim 20, characterized in that The method further comprises: The third information is respectively sent to the M third network elements, wherein the third information sent to one third network element among the M third network elements indicates features for participating in the training of the distributed learning model.

22. The method according to claim 20 or 21, characterized in that Determining at least one third network element includes: receiving fourth information from the second network element, where the fourth information indicates information of the at least one third network element; Determine the at least one third network element according to the fourth information.

23. A communication method, characterized in that: Applied to a third network element, the method includes: receiving first information from a first network element, where the first information is used to request features supported by the third network element; Second information is sent to the first network element, where the second information indicates features supported by the third network element.

24. The method according to claim 23, wherein Receive third information from the first network element, wherein the third information indicates characteristics of the third network element for participating in distributed learning model training.

25. The method according to claim 23 or 24, characterized in that The method further comprises: Fifth information is sent to the second network element, where the fifth information indicates information of the third network element.

26. A communication device, characterized in that: include: A module for executing the method according to any one of claims 1 to 11; or A module for performing the method according to any one of claims 12 to 19; or A module for performing the method according to any one of claims 20 to 22; or A module for executing the method according to any one of claims 23 to 25.

27. A communication device, characterized in that: include: A processor and a memory, wherein the memory is used to store program instructions, and the processor is used to execute the program instructions in the memory to implement the method according to any one of claims 1 to 11, the method according to any one of claims 12 to 19, the method according to any one of claims 20 to 22, or the method according to any one of claims 23 to 25.

28. A communication device, characterized in that: include: Processing circuit and interface circuit; wherein: The interface circuit is used to couple with a memory outside the communication device and provide a communication interface for the processing circuit to access the memory; the processing circuit is used to execute program instructions in the memory to implement the method according to any one of claims 1 to 11, the method according to any one of claims 12 to 19, the method according to any one of claims 20 to 22, or the method according to any one of claims 23 to 25.

29. A computer program product comprising instructions, characterized in that When the instruction is executed by the communication device, the communication device executes the method according to any one of claims 1-11, implements the method according to any one of claims 12-19, implements the method according to any one of claims 20-22, or implements the method according to any one of claims 23-25.

30. A computer-readable storage medium, characterized in that The storage medium stores a computer program or instruction. When the computer program or instruction is executed by the communication device, it implements the method according to any one of claims 1 to 11, the method according to any one of claims 12 to 19, the method according to any one of claims 20 to 22, or the method according to any one of claims 23 to 25.

Images