Client discovery method and related equipment

Abstract

The invention provides a client discovery method and related equipment, and relates to the technical field of communication. The method is executed by an inference longitudinal federation learning server. The method comprises the following steps: acquiring a training longitudinal federated learning client network element identifier from a training longitudinal federated learning server; and finding the matched inference longitudinal federated learning client network element according to the training longitudinal federated learning client network element identifier.

Description

Technical Field

[0001] This disclosure relates to the field of communication technology, and in particular to a client discovery method, apparatus, communication device, computer-readable storage medium, and computer program product. Background Technology

[0002] Some communication systems (e.g., 5G) have introduced artificial intelligence (AI) capabilities, enabling models to process network services. However, these technologies only support independent learning by network nodes to acquire models. This can lead to situations where network nodes may be unable to train models in certain scenarios (such as when there is insufficient training data), or result in poor model training performance. Summary of the Invention

[0003] This disclosure provides a client discovery method executed by an inference longitudinal federated learning server. The method includes: obtaining a training longitudinal federated learning client network element identifier from a training longitudinal federated learning server; and discovering a matching inference longitudinal federated learning client network element based on the training longitudinal federated learning client network element identifier. In exemplary embodiments, the method may be executed by the inference longitudinal federated learning server, or the method may be executed by a device (e.g., a chip) configured on the inference longitudinal federated learning server.

[0004] This disclosure provides a client discovery method, executed by a training longitudinal federated learning server. The method includes: enabling the inference longitudinal federated learning server to obtain the network element identifier of the training longitudinal federated learning client. In exemplary embodiments, the method can be executed by the training longitudinal federated learning server, or the method can be executed by a device (e.g., a chip) configured on the training longitudinal federated learning server.

[0005] This disclosure provides a client discovery method executed by a network storage function network element. The method includes: receiving a network element discovery request sent by an inference longitudinal federated learning server, the network element discovery request including a training longitudinal federated learning client network element identifier; searching for a matching inference longitudinal federated learning client network element identifier based on the network element discovery request; and sending the matching inference longitudinal federated learning client network element identifier to the inference longitudinal federated learning server. In exemplary embodiments, this method can be executed by a network storage function network element, or by a device (e.g., a chip) configured in the network storage function network element.

[0006] This disclosure provides a client discovery method, executed by an analysis logic function network element, comprising: sending a registration request to a network storage function network element. The registration request includes network element configuration of the analysis logic function network element, the network element configuration including at least one of network data analysis function network element type, analysis identifier, network data analysis function address information, service range, and vertical federated learning capability information; the vertical federated learning capability information includes a corresponding model training logic function identifier, which indicates a model training logic function identifier configured internally within the analysis logic function network element; or, indicates a model training logic function identifier configured internally within the analysis logic function network element that can be used for vertical federated learning. In an exemplary embodiment, the method can be executed by the analysis logic function network element, or the method can be executed by a device (e.g., a chip) configured in the analysis logic function network element.

[0007] This disclosure provides a client discovery device applied to an inference longitudinal federated learning server. The device includes: an acquisition unit for acquiring a training longitudinal federated learning client network element identifier from the training longitudinal federated learning server; and a processing unit for discovering a matching inference longitudinal federated learning client network element based on the training longitudinal federated learning client network element identifier.

[0008] This disclosure provides a client discovery device applied to a training longitudinal federated learning server. The device includes a processing unit configured to enable the inference longitudinal federated learning server to obtain the network element identifier of the training longitudinal federated learning client.

[0009] This disclosure provides a client discovery device applied to a network storage function network element. The device includes: a receiving unit for receiving a network element discovery request sent by an inference vertical federated learning server, the network element discovery request including a training vertical federated learning client network element identifier; a processing unit for searching for a matching inference vertical federated learning client network element identifier according to the network element discovery request; and a sending unit for sending the matching inference vertical federated learning client network element identifier to the inference vertical federated learning server.

[0010] This disclosure provides a client discovery device applied to an analysis logic function network element. The device includes a sending unit for sending a registration request to a network storage function network element. The registration request includes network element configuration of the analysis logic function network element, which includes at least one of the following: network data analysis function network element type, analysis identifier, network data analysis function address information, service range, and vertical federated learning capability information. The vertical federated learning capability information includes a corresponding model training logic function identifier, which indicates a model training logic function identifier configured within the analysis logic function network element; or, indicates a model training logic function identifier configured within the analysis logic function network element that can be used for vertical federated learning.

[0011] This disclosure provides a communication system, which includes: an inference longitudinal federated learning server as described in any embodiment of this disclosure, a training longitudinal federated learning server as described in any embodiment of this disclosure, a network storage function network element as described in any embodiment of this disclosure, and an analysis logic function network element as described in any embodiment of this disclosure.

[0012] This disclosure provides a communication device. In one design, the communication device may include modules / units that perform the methods / operations / steps / actions described in any embodiment of this disclosure. These modules / units may be hardware circuits, software, or a combination of hardware circuits and software.

[0013] This disclosure provides a communication device, including a processor. The processor can implement the methods in any embodiment of this disclosure. Optionally, the communication device further includes a memory, and the processor is coupled to the memory and can be used to execute a computer program stored in the memory to implement the methods in any embodiment of this disclosure. Optionally, the communication device further includes a communication interface, and the processor is coupled to the communication interface. In this disclosure, the communication interface can be a transceiver, an input / output interface, a pin, a circuit, a bus, a module, or other types of communication interface, and is not limited thereto.

[0014] Optionally, the transceiver can be a transceiver circuit. Optionally, the input / output interface can be an input / output circuit.

[0015] This disclosure also provides a processor, including an input circuit, an output circuit, and a processing circuit. The processing circuit is used to receive signals through the input circuit and to transmit signals through the output circuit, causing the processor to execute the method in any embodiment of this disclosure.

[0016] In exemplary embodiments, the processor described above can be one or more chips, the input circuit can be an input pin, the output circuit can be an output pin, and the processing circuit can be a transistor, a gate circuit, a flip-flop, and various logic circuits, etc. The input signal received by the input circuit can be received and input by, for example, but not limited to, a receiver, and the signal output by the output circuit can be, for example, but not limited to, output to a transmitter and transmitted by the transmitter. Furthermore, the input circuit and the output circuit can be the same circuit, which is used as both the input circuit and the output circuit at different times. This disclosure does not limit the specific implementation of the processor and various circuits.

[0017] This disclosure also provides a communication system, including at least one communication device as described in this disclosure.

[0018] This disclosure provides a communication device, which includes: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute a client discovery method according to any embodiment of this disclosure by executing the executable instructions.

[0019] This disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the client discovery method in any embodiment of this disclosure.

[0020] This disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the client discovery method in any embodiment of this disclosure. Attached Figure Description

[0021] Figure 1 A flowchart illustrating a client discovery method according to an embodiment of this disclosure is shown.

[0022] Figure 2 This diagram illustrates a client discovery method according to an embodiment of the present disclosure.

[0023] Figure 3 A schematic diagram of yet another client discovery method in an embodiment of this disclosure is shown.

[0024] Figure 4 A schematic diagram of another client discovery method in an embodiment of this disclosure is shown.

[0025] Figure 5 A schematic diagram of another client discovery method in an embodiment of this disclosure is shown.

[0026] Figure 6 A flowchart illustrating another client discovery method in an embodiment of this disclosure is shown.

[0027] Figure 7A flowchart of yet another client discovery method according to an embodiment of this disclosure is shown.

[0028] Figure 8 A flowchart of yet another client discovery method according to an embodiment of this disclosure is shown.

[0029] Figure 9 A schematic diagram of a client discovery device according to an embodiment of this disclosure is shown.

[0030] Figure 10 A schematic diagram of another client discovery device is shown in an embodiment of this disclosure.

[0031] Figure 11 A schematic diagram of yet another client discovery device is shown in an embodiment of this disclosure.

[0032] Figure 12 A schematic diagram of another client discovery device is shown in an embodiment of this disclosure.

[0033] Figure 13 A structural block diagram of a communication device according to an embodiment of the present disclosure is shown. Detailed Implementation

[0034] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0035] Furthermore, the accompanying drawings are merely illustrative of this disclosure, and the same reference numerals in the drawings denote the same or similar parts, thus repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0036] The following explains some concepts and / or terms involved in the embodiments of this disclosure.

[0037] 5GC (5Generation Core): 5G core network.

[0038] RAN (Radio Access Network): Wireless access network.

[0039] AMF (Access and Mobility Management Function): Access and Mobility Management Function / Access and Mobility Management Network Element / Access and Mobility Management Entity.

[0040] NF (Network Function): Network element / entity.

[0041] AnLF (Analytics Logical Function): An analytical logical function / analytical logical function network element / analytical logical function entity, a type of NWDAF, sometimes also written as ANLF below.

[0042] MTLF (Model Training Logical Function): A type of NWDAF (Model Training Logical Function Network Element / Model Training Logical Function Entity).

[0043] ADRF (Analytics Data Repository Function): Analytical data storage function.

[0044] UE (User Equipment): User terminal, user equipment, terminal.

[0045] NWDAF (Network Data Analytics Function): Network data analytics function / network data analytics function element / network data analytics function entity. The NWDAF element can be decomposed into two parts: NWDAF (AnLF) and NWDAF (MTLF). The former is the network element responsible for the inference function (AnLF), and the latter is the network element responsible for the training function (MTLF).

[0046] HFL (Horizontal Federated Learning): Horizontal federated learning.

[0047] VFL (Vertical Federated Learning): Vertical federated learning.

[0048] Artificial Intelligence (AI) and AI Models: Artificial intelligence has been widely applied in various fields. AI models have various algorithmic implementations, such as neural networks, decision trees, support vector machines, and Bayesian classifiers. This disclosure uses neural networks as an example for illustration, but does not limit the specific type of AI module. A neural network consists of neurons; an input layer, hidden layers, and an output layer composed of numerous neurons constitute a neural network. The number of hidden layers and the number of neurons in each layer constitute the "network structure" of the neural network. The parameter information of each neuron, combined with the algorithm used, constitutes the "parameter information" of the entire network, which is also an important part of the AI ​​model file. In practical use, an AI model refers to a file containing elements such as network structure and parameter information. A trained AI model can be directly reused by its framework platform without repeated construction or learning, directly performing intelligent functions such as judgment and recognition.

[0049] The specific implementation methods of the embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. The solutions provided in the embodiments of this disclosure relate to wireless communication and terminal technologies.

[0050] Figure 1 A flowchart illustrating a client discovery method according to an embodiment of this disclosure is shown. Figure 1 The method provided in the embodiments can be executed by the inference longitudinal federated learning server.

[0051] In this embodiment, vertical federated learning differs from horizontal federated learning techniques. Vertical federated learning uses the same data samples but has different sample features. For example, in 5GC, for a user's QoE (Quality of Experience) information, the data in AF (Application function / Application function network element / Application function entity) and NWDAF have different data features / sample features and the same data samples, respectively. Vertical federated learning in 5GC can effectively break down data isolation and obtain a better-performing model / AI model / vertical federated learning model / VFL model by exchanging intermediate data such as gradients and losses.

[0052] In this embodiment, the AI ​​function of 5GC is divided into two processes: training and inference, which can be performed by MTLF and ANLF respectively. MTLF is responsible for collecting a large amount of data to train the model, while ANLF, after subscribing to the model from MTLF, is only responsible for collecting a small amount of data for inference analysis.

[0053] In this embodiment of the disclosure, during vertical federated learning, a VFL server (MTLF or AF) initiates one or more clients(s) (multiple MTLFs or AFs) to collaboratively complete the vertical federated learning training process. Ultimately, the VFL server and VFL client(s) each hold their trained local models, enabling model training to be completed without data leaving the network element. The VFL server that collaboratively completes the vertical federated learning training process is referred to as the training vertical federated learning server or training vertical federated learning client, hereinafter also referred to as the VFL server in the training phase; the VFL client(s) are referred to as the training vertical federated learning client or training vertical federated learning client network element, hereinafter also referred to as the VFL client(s) in the training phase. The VFL server and VFL client(s) in the training phase can be network nodes used for model training, and therefore can also be referred to as model training functional nodes.

[0054] In this scenario, the inference process will be similar to the training process, with one VFL server (ANLF or AF) initiating one or more clients(s) (multiple ANLFs or AFs) to jointly complete the inference process. The VFL server that jointly participates in the inference process in vertical federated learning is called the inference vertical federated learning server or inference vertical federated learning server, hereinafter also referred to as the VFL server in the inference phase; the VFL clients(s) are called inference vertical federated learning clients or inference vertical federated learning client network elements, hereinafter also referred to as VFL client(s) in the inference phase. The VFL server and VFL client(s) in the inference phase can be network nodes used for inference to generate prediction information, generate statistical information, or perform data analysis, and therefore can also be called model inference function nodes.

[0055] In some embodiments, the model training functional node can be a network element, terminal, or module in the communication network that has AI model training capabilities; the model inference functional node can be a network element, terminal, or module in the communication network that has model inference capabilities. That is to say, the model training functional node and the model inference functional node can also be called by other names.

[0056] In this embodiment, the model training and model inference nodes can be core network elements or internal modules of core network elements. For example, NWDAF can include model training and model inference nodes, VFLserver is a core network element with vertical federated learning server capabilities, and VFL client(s) can be other core network elements or modules participating in vertical federated learning. Alternatively, the model training and model inference nodes can be radio access network elements or internal modules of radio access network elements, that is, the model training and model inference nodes can be functions within the RAN. For example, the model training node can be a RAN device with model training capabilities, specifically a base station device or module with model training capabilities, the VFL server is a base station device or module with vertical federated learning server capabilities or Operation Administration and Maintenance (OAM), and the VFL client(s) can be other member base station devices or modules participating in vertical federated learning.

[0057] In some embodiments, the model training function node and the model inference function node can also be functions of the terminal or within the terminal. For example, the model training function node can be a terminal with model training function, the VFL server can be a terminal or AF with vertical federated learning server capability, and the VFL client(s) can be other member terminals participating in vertical federated learning.

[0058] It should be noted that in this embodiment, the model training function node and the model inference function node can be deployed independently as different network element devices, or co-deployed in the same network element device, such as two internal functional modules of the core network element, the wireless access network element, or the terminal. In this case, the core network element, the wireless access network element, or the terminal can provide both AI model training function and model inference function.

[0059] In this embodiment, the VFL server is a network element that coordinates or hosts the vertical federated learning process. For example, the VFL server can be a VFL central network element or a VFL coordinator network element, or a central model training function node used for VFL operations. The VFL client(s) can specifically be a core network element, a radio access network element, a terminal, or an application server, etc. The VFL client(s) are network elements participating in the vertical federated learning process, and can be referred to as network elements participating in VFL operations.

[0060] In some embodiments, the model training function node can be MTLF, and the model inference function node can be AnLF.

[0061] In the above scenario, the VFL server during the inference phase needs to ensure that the VFL client(s) participating in inference is either an AF that has participated in training, or an ANLF corresponding to an MTLF that has participated in training, when it discovers VFL client(s) during the inference phase. However, related technologies can only require requirements such as the capabilities and service scope of the ANLF when searching for it, and cannot associate it with its corresponding MTLF.

[0062] like Figure 1 As shown, the method provided in this disclosure embodiment may include the following steps.

[0063] In S110, the network element identifier of the training longitudinal federated learning client (i.e., the NF ID of the VFL client(s) during the training phase) is obtained from the training longitudinal federated learning server (i.e., the VFL server during the training phase).

[0064] In some embodiments, the VFL server during the training phase can be the same as the VFL server during the inference phase, for example, the same Open Field Algorithm (AF). In other embodiments, the VFL server during the training phase can be different from the VFL server during the inference phase; for example, one may be an MTLF and the other an AnLF. In this case, the VFL server during the training phase can send the NF IDs of the VFL clients during the training phase to the VFL server during the inference phase, and the VFL server during the inference phase can receive the NF IDs of the VFL clients during the training phase from the VFL server during the training phase.

[0065] In an exemplary embodiment, the network element identifier for the training longitudinal federated learning client includes at least one of the following:

[0066] Model training logic function identifier (i.e. MTLF ID(s) that participate in training or training phase) so that the VFL Server in the inference phase can discover the ANLF corresponding to the MTLF ID(s) in the training phase, so as to serve as the VFL client(s) in the inference phase.

[0067] The application uses functional identifiers (i.e., AF ID(s) that participate in training or the training phase) so that the VFL Server in the inference phase can directly initiate VFL inference to the AF corresponding to the AF ID(s) based on the AF ID(s) in the training phase.

[0068] In some embodiments, MTLF(s) may be referred to as MTLF network element, NWDAF network element, or NWDAF containing MTLF network element.

[0069] In some embodiments, the training of the vertical federated learning server determines the network element identifier of the training vertical federated learning client, including: the training of the vertical federated learning server sending a node discovery request message to the network repository function network element, the node discovery request message being used to request network nodes participating in the vertical federated learning training; the training of the vertical federated learning server receiving a response message sent by the network repository function network element, the response message including information of the VFL client(s) in the training phase (e.g., the NF ID of the VFL client(s) in the training phase).

[0070] In this embodiment of the disclosure, the network repository function element can store information of one or more VFL client(s), so that after receiving the above-mentioned node discovery request message, it returns the information of the VFL client(s) of the corresponding training stage to the above-mentioned training longitudinal federated learning server.

[0071] In this embodiment of the disclosure, the network warehouse function network element is a network element responsible for capacity storage. This network element can be at least one of NRF network element, UDM (User Data Management) network element, Data Collection Application Function (DCAF) network element, etc.

[0072] In this implementation, information about the VFL client(s) during the training phase can be obtained through the network repository function element. It should be noted that in some implementations, the information about the VFL client(s) during the training phase may not need to be obtained from the network repository function element; for example, the information about the VFL client(s) during the training phase may be fixedly configured in the training longitudinal federated learning server.

[0073] Optionally, the node discovery request message may include at least one of the following: Analytics ID, Area of ​​Interest (AOI) information, Time of Interest information, Model description information, Model shareability information, Model performance information, Model algorithm information, Model training speed information, Federated learning instruction information, Federated learning type information, VFL server node type instruction information, VFL client node type instruction information, First service information, and Second service information.

[0074] The aforementioned analysis identifier can be used to indicate that the network node requested by the node discovery request message needs to support the model training task corresponding to the analysis identifier.

[0075] The aforementioned AOI information can be used to indicate that the network node requested by the node discovery request message can serve the area corresponding to the AOI information, which may be at least one Tracking Area (TA), at least one cell, or other areas.

[0076] The aforementioned federated learning instruction information can be used to indicate that the network node requested by the node discovery request message needs to support federated learning.

[0077] The above federated learning type information can be used to indicate that the type of federated learning that the network node requested by the node discovery request message needs to support is the vertical federated learning type.

[0078] Among them, the horizontal federated learning type can be to use different training data samples with the same feature points for learning and training; the vertical federated learning type can be to use training data samples with different feature points of the same training sample for learning and training.

[0079] In S120, matching inference longitudinal federated learning client elements (i.e., VFL client(s) in the inference phase) are discovered based on the training longitudinal federated learning client element identifier.

[0080] The method provided in this disclosure allows the VFL server in the inference phase to obtain the NF IDs of the VFL clients(s) participating in the training phase from the VFL server in the training phase. This enables the VFL server in the inference phase to discover VFL clients(s) corresponding to the NF IDs of the VFL clients(s) participating in the training phase, and to use these VFL clients(s) as VFL clients(s) in the inference phase. This improves the discovery efficiency and accuracy of VFL clients(s) in the inference phase.

[0081] In an exemplary embodiment, discovering a matching inference longitudinal federated learning client network element based on the training longitudinal federated learning client network element identifier includes: sending a network element discovery request to a network storage function (NRF) network element, the network element discovery request including the training longitudinal federated learning client network element identifier; and receiving the inference longitudinal federated learning client network element identifier returned by the network storage function network element in response to the network element discovery request, the inference longitudinal federated learning client network element identifier corresponding to the inference longitudinal federated learning client network element.

[0082] In this embodiment of the disclosure, the mapping relationship between each ANLF and its corresponding MTLF can be pre-stored in the NRF. When the VFL server in the inference phase obtains the NF ID of the VFL client(s) participating in training from the VFL server in the training phase, if the NF ID indicates that the VFL client(s) participating in training is an MTLF instead of an AF (i.e., the NF ID is an MTLF ID(s)), it can send a network element discovery request to the NRF, which carries the NF ID of the VFL client(s) participating in training. When the NRF receives the network element discovery request, it can match the NF ID of the VFL client(s) participating in training with the pre-stored mapping relationship between each ANLF and its corresponding MTLF according to the NF ID of the VFL client(s) participating in training carried in the network element discovery request. The ANLF ID(s) corresponding to the MTLF that matches the NF ID of the VFL client(s) participating in training is taken as the VFL client(s) in the inference phase and returned to the VFL server in the inference phase.

[0083] In other embodiments, when the VFL server in the inference phase obtains the NF ID of the VFL client(s) participating in the training from the VFL server in the training phase, if the NF ID indicates that the AF is participating in the training instead of the MTLF (i.e., the NF ID is the AF ID(s)), then the AF participating in the training can be directly used as the VFL client(s) in the inference phase.

[0084] In some embodiments, the aforementioned network element discovery request / network element discovery request message may also include other information, such as service location information, filtering information, etc., which are not limited in this disclosure.

[0085] In an exemplary embodiment, the network element discovery request includes at least one of the following:

[0086] Vertical federated learning inference instruction information, which is used to indicate that the inference vertical federated learning client network element needs to support the vertical federated learning inference function and be able to participate in the vertical federated learning inference process;

[0087] Vertical federated learning capability indication information, which is used to indicate that the inference vertical federated learning client network element needs to support the vertical federated learning function and be able to participate in the vertical federated learning process;

[0088] Vertical federated learning client capability indication information, which is used to indicate that the inference vertical federated learning client network element needs to be able to participate in the vertical federated learning process as a vertical federated learning client.

[0089] The model training logic function identifier is used to indicate that the model training logic function network element corresponding to the inference longitudinal federated learning client network element contains the model training logic function network element specified by the training longitudinal federated learning client network element identifier.

[0090] Network data analysis function network element types;

[0091] Analysis identifiers;

[0092] Service scope.

[0093] For example, during the inference phase, the VFL Server (i.e., the server mentioned above) sends a network element discovery request to the NRF, which includes at least one of the following:

[0094] VFL inference (Vertical Federated Learning Inference Instruction Information): Indicates that the inference client element needs to support VFL inference functionality and be able to participate in the VFL inference process;

[0095] VFL Capability (Vertical Federated Learning Capability Indication Information): Indicates that the inference client network element needs to support VFL functionality and be able to participate in the VFL process;

[0096] VFL client capability (Vertical Federated Learning Client Capability Indication Information): Indicates that the inference client network element needs to be able to participate in the VFL process as a VFL client;

[0097] MTLF ID(s) (Model Training Logic Function Identifier): Indicates that the MTLF corresponding to the inference client network element needs to contain a specified MTLF (the specified MTLF is the MTLF corresponding to the MTLF ID(s) carried in the network element discovery request, i.e. the MTLF in the training phase mentioned above).

[0098] NWDAF NF type (Network Data Analysis Function Network Element Type);

[0099] Analytics ID(s);

[0100] Service Area.

[0101] Based on the information carried in the aforementioned network element discovery request, NRF finds the inference client network element that satisfies the information in the aforementioned network element discovery request and returns it as the matching inference vertical federated learning client network element to the inference vertical federated learning server.

[0102] In one optional implementation, the model inference-related capability information of the inference client network element (such as the aforementioned VFLinference, VFL capability, VFL client capability, MTLF ID(s)) corresponds to the Analytics ID(s), and can be indicated by each Analytics ID, with different Analytics IDs corresponding to different capabilities.

[0103] In an exemplary embodiment, the training longitudinal federated learning server is a model training logic function network element. In an exemplary embodiment, the training longitudinal federated learning client may be one of the following: a model training logic function network element, or an application function network element, or multiple model training logic function network elements, or multiple model training logic function network elements and one application function network element, or multiple application function network elements, or multiple model training logic function network elements and multiple application function network elements.

[0104] In an exemplary embodiment, the inference vertical federated learning server is an analysis logic function network element. In an exemplary embodiment, the inference vertical federated learning client network element is one of the following: one analysis logic function network element, or one application function network element, or multiple analysis logic function network elements, or multiple analysis logic function network elements and one application function network element, or multiple application function network elements, or multiple analysis logic function network elements and multiple application function network elements.

[0105] In an exemplary embodiment, the method provided in this disclosure further includes: receiving an analysis identifier request sent by a consumer network element as an analysis logic function network element serving as the inference vertical federated learning server; responding to the analysis identifier request, discovering a model training logic function network element that can serve as a training vertical federated learning server from a network storage function network element; and sending a model acquisition request to the model training logic function network element serving as the training vertical federated learning server to request the acquisition of the model for vertical federated learning.

[0106] In an exemplary embodiment, the method provided in this disclosure further includes: receiving information about a trained model from a model training logic function network element that serves as a training longitudinal federated learning server.

[0107] In an exemplary embodiment, the information of the trained model is the model file and / or the model's download address.

[0108] In an exemplary embodiment, the method provided in this disclosure further includes: the inference vertical federated learning server and the inference vertical federated learning client network element completing the inference process and obtaining the inference result.

[0109] In an exemplary embodiment, the training vertical federated learning server is an application function network element.

[0110] In an exemplary embodiment, the training vertical federated learning client is one of the following: a model training logical function network element, or multiple model training logical function network elements.

[0111] In an exemplary embodiment, the inference vertical federated learning server is an application function network element.

[0112] In an exemplary embodiment, the inference vertical federated learning client network element is one of the following: a single analysis logic function network element, or multiple analysis logic function network elements.

[0113] In an exemplary embodiment, the method provided in this disclosure further includes: receiving a vertical federated learning request sent by an analysis logic function network element through the application function network element; responding to the vertical federated learning request by discovering a training vertical federated learning client from a network storage function network element; and performing vertical federated learning training with the training vertical federated learning client through the application function network element to obtain a trained model.

[0114] In an exemplary embodiment, the application function network element uses the training longitudinal federated learning client identifier of the training longitudinal federated learning client to discover a matching inference longitudinal federated learning client network element.

[0115] In an exemplary embodiment, the network element identifier for the training longitudinal federated learning client includes: the identifier of the model training logic function participating in the training.

[0116] In an exemplary embodiment, the method provided by this disclosure further includes: completing the inference process through the application function network element and the inference vertical federated learning client network element to obtain the inference result; and returning the inference result to the analysis logic function network element that sent the vertical federated learning request.

[0117] In an exemplary embodiment, discovering a matching inference longitudinal federated learning client element based on the training longitudinal federated learning client element identifier includes: sending a longitudinal federated learning inference request to the candidate analysis logic function element; receiving response information returned by the candidate analysis logic function element in response to the longitudinal federated learning inference request; and determining the inference longitudinal federated learning client element from the candidate analysis logic function elements based on the response information.

[0118] In an exemplary embodiment, the method provided in this disclosure further includes: determining the candidate analysis logic function network element from the network storage function network element.

[0119] In an exemplary embodiment, the vertical federated learning inference request includes the identifier of the training vertical federated learning client network element, which includes the identifier of the model training logic function; the response information includes indication information on whether the corresponding candidate analysis logic function network element is added to the vertical federated learning inference process.

[0120] In an exemplary embodiment, the candidate analysis logic function network element is used to determine whether to participate in longitudinal federated learning inference based on the training longitudinal federated learning client network element identifier, so as to generate the indication information.

[0121] In an exemplary embodiment, the candidate analysis logic function network element is used to determine whether to participate in longitudinal federated learning inference based on at least one of the following:

[0122] The identifier of the training longitudinal federated learning client network element is compared with the identifier of the model training logic function configured inside the candidate analysis logic function network element in turn to determine whether they match.

[0123] The network storage function element discovers model training logic function elements that can provide services to the candidate analysis logic function elements. The identifier of the training longitudinal federated learning client element and the identifier of the model training logic function returned by the network storage function element are compared in turn to determine whether they match.

[0124] In this embodiment of the disclosure, ANLF can also be referred to as AnLF network element, NWDAF network element, or NWDAF containing AnLF network element.

[0125] In an exemplary embodiment, discovering a matching inference longitudinal federated learning client network element based on the training longitudinal federated learning client network element identifier includes: when the training longitudinal federated learning client network element identifier is an application function identifier, then determining the application function network element corresponding to the application function identifier as the inference longitudinal federated learning client network element.

[0126] In some embodiments, after the VFL server completes training during the training phase, it informs the VFL server (e.g., ANLF) during the inference phase of the VFL client network element IDs that participated in the training. These VFL client network element IDs include AFIDs, so that the VFL server can directly initiate VFL inference to the AF corresponding to the AF ID.

[0127] In this embodiment of the disclosure, the ANLF can find its corresponding MTLF in the following two ways, but this disclosure is not limited to these:

[0128] (1) ANLF finds its corresponding MTLF(s) based on the MTLF ID(s) configured internally by the operator;

[0129] (2) ANLF uses NRF (Network Repository Function) to find a suitable MTLF(s).

[0130] Based on the above, this disclosure primarily addresses the issue in vertical federated learning scenarios where, due to the inconsistency between the network elements responsible for training and inference in the core network, the VFL server during the inference process / stage needs to be able to find the inference network elements (i.e., VFL clients in the inference stage) corresponding to the VFL clients during the training process to complete the inference process. To discover inference clients (i.e., inference network elements) that meet the above requirements, this disclosure proposes that the ANLF registers its internally configured MTLF ID(s) with the NRF. Simultaneously, the VFL server during the training stage informs the VFL server during the inference stage of the NF IDs (network element identifiers, i.e., training vertical federated learning client network element identifiers) of the VFL clients participating in training, so that it can use these NF IDs to discover the corresponding inference clients, such as the ANLF. This disclosure proposes a corresponding solution to help the VFL server during the inference process find suitable VFL clients for the inference stage. The following is a related explanation. Figure 2 Examples are provided, but this disclosure is not limited thereto.

[0131] Figure 2 This diagram illustrates a client discovery method according to an embodiment of the present disclosure. Figure 2 As shown, the method provided in this disclosure embodiment may include the following steps.

[0132] In S21, when ANLF registers with NRF, it also needs to include its internally configured MTLF ID.

[0133] In this embodiment of the disclosure, the ANLF sends a registration request to the NRF, that is, the ANLF registers the NF profile (Network Element Configuration / Network Element Information) with the NRF. The registration request includes at least one of the following: NWDAF NF type, Analytics ID(s), Address information of NWDAF, Service Area, VFL capability information, etc.

[0134] In an exemplary embodiment, the registration request includes VFL capability information, which contains at least one of the following:

[0135] VFL inference (Vertical Federated Learning Inference Instruction Information): Indicates that the ANLF supports VFL inference functionality and can participate in the VFL inference process;

[0136] VFL Capability (Vertical Federated Learning Capability Indicator): Indicates that the ANLF supports VFL functionality and is able to participate in the VFL process;

[0137] VFL server capability (Vertical Federated Learning Server Capability Indicator Information): Indicates that the ANLF can participate in the VFL process as a VFL server;

[0138] VFL client capability (Vertical Federated Learning Client Capability Indicator Information): Indicates that the ANLF can participate in the VFL process as a VFL client;

[0139] The corresponding MTLF ID(s) (corresponding model training logic function identifier): indicates the MTLF ID(s) configured internally by this ANLF. When this ANLF needs to interact with an MTLF, it sends a request to the corresponding MTLF based on this MTLF ID(s). Alternatively, it indicates the MTLF ID(s) configured internally by this ANLF that can be used for VFL.

[0140] It should be noted that since not all MTLFs can execute VFLs, "MTLF ID(s) that are configured internally within the ANLF and can be used for VFLs" refers to MTLF ID(s) specifically configured for VFLs, while "MTLF ID(s) that are configured internally within the ANLF" refers to ordinary MTLFs.

[0141] In related technologies, core network element ANLFs only register information related to their own capabilities and do not register their internally configured MTLF information. This application allows ANLFs to also register their internally configured MTLF information with the NRF, thereby enabling the discovery of ANLFs corresponding to the MTLFs during the training phase using the MTLF information stored in the NRF, which can then be used as VFL clients during the inference phase.

[0142] In S22, the VFL Server in the training phase informs the VFL Server in the inference phase of the NF IDs of the client network elements (i.e., the VFL client(s)) participating in the training phase.

[0143] In some embodiments, after the server in the training phase completes training, it informs the server in the inference phase (e.g., ANLF) of the client network element IDs that participated in the training (i.e., the NF IDs of the client network elements that participated in the training phase).

[0144] In an exemplary embodiment, the client network element IDs participating in the training include at least one of the following:

[0145] MTLF ID(s) (i.e. MTLF ID(s) used in training): so that the VFL Server in the inference phase can discover the ANLF corresponding to the MTLF of the MTLF ID(s) and use it as the VFL client(s) in the inference phase.

[0146] AF ID (i.e., the AF ID used in training): so that the VFL Server in the inference phase can directly initiate VFL inference to the AF corresponding to this AF ID.

[0147] In this embodiment of the disclosure, if the client used in the training phase is AF, then the client used in the inference phase is also the same AF.

[0148] In S23, the VFL Server in the inference phase uses the client network element ID from the training phase (i.e., the client network element participating in the training) to discover the client network element in the inference phase.

[0149] In this embodiment of the disclosure, the VFL Server during the inference phase sends a network element discovery request to the NRF. The network element discovery request includes at least one of the following: VFL inference; VFL capability; VFL client capability; MTLF ID(s); NWDAF NFtype; Analytics ID(s); Service Area.

[0150] In S24, NRF returns a satisfactory inference phase client to the VFL Server during the inference phase.

[0151] In this embodiment of the disclosure, the NRF searches for matching network elements based on the information in the received network element discovery request, identifies them as clients for the inference phase, and returns them to the VFL Server for the inference phase. For example, the ID of the matching network element can be returned to the VFL Server for the inference phase to serve as a VFL client for the inference phase.

[0152] In an exemplary embodiment, the NRF network element discovery method includes at least one of the following:

[0153] (1) The NRF determines that the VFL inference in the network element discovery request matches the information in the registered NF profile. For example, both the VFL inference in the network element discovery request and the VFL inference in the NF profile indicate that the VFL inference function is supported and can participate in the VFL inference process.

[0154] (2) The NRF determines that the VFL capability in the network element discovery request matches the information in the registered NF profile. For example, both the VFL capability in the network element discovery request and the VFL capability in the NF profile indicate that the VFL function is supported and that the network element can participate in the VFL process.

[0155] (3) The NRF determines that the VFL client capabilities in the network element discovery request match the information in the registered NF profile. For example, both the VFL client capabilities in the network element discovery request and the VFL client capabilities in the NF profile indicate that the client can participate in the VFL process as a VFL client.

[0156] (4) The NRF determines that the MTLF ID(s) in the network element discovery request matches the information in the registered NF profile. For example, the MTLF ID(s) in the network element discovery request and the MTLF ID(s) in the NF profile can match.

[0157] (5) The NRF determines that the NWDAF NF type in the network element discovery request matches the information in the registered NF profile, for example, the NWDAF NF type in the network element discovery request matches the NWDAF NF type in the NF profile;

[0158] (6) The NRF determines that the Analytics ID(s) in the network element discovery request matches the information in the registered NF profile, for example, the Analytics ID(s) in the network element discovery request matches the Analytics ID(s) in the NF profile;

[0159] (7) The NRF determines that the Service Area in the network element discovery request matches the information in the registered NF profile, for example, the Service Area in the network element discovery request matches the Service Area in the NF profile.

[0160] It is understood that this disclosure does not limit the execution order of S21 and S22; they can be executed sequentially or in parallel.

[0161] In related technologies, finding ANLF network elements only requires providing information such as analytics ID and AOI; NRF cannot find ANLFs using MTLF IDs. Addressing the issue of inaccurate VFL client discovery in vertical federated learning in 5GC, this disclosure proposes that the ANLF registers its internally configured MTLF ID with the NRF. During the training phase, the server informs the inference phase server of the NF IDs of the clients participating in training, enabling the inference phase server to discover the corresponding inference client using the NF ID of the client participating in training, such as the ANLF corresponding to the MTLF participating in training, and thus use it as the inference phase client. This disclosure provides a client discovery method supporting multi-party collaborative inference in a vertical federated learning scenario within 5GC.

[0162] The solution proposed in this disclosure supports the inference process in vertical federated learning between network elements and application servers in 5GC. It is a key technology supporting vertical federated learning, breaking down data isolation and effectively solving the data silo problem, enabling better completion of AI tasks without exposing data. The solution proposed in this disclosure addresses the data silo problem faced during the training and inference of AI / ML (Machine Learning) models. While avoiding the exposure of network data by operators, it completes the model inference process, forming a superior AI / ML model, thereby providing better services.

[0163] On the one hand, the solution proposed in this disclosure enables the server in the VFL inference phase to efficiently find its corresponding VFL client, effectively reducing signaling interactions. On the other hand, the solution proposed in this disclosure reuses and enhances the network architecture, requiring minimal modification to the overall system while meeting functional requirements, and is scalable. Furthermore, the enhanced functions of the solution proposed in this disclosure are mainly on the core network side and the application side, requiring no additional processing from the UE, minimizing the impact on users, and providing a better user experience.

[0164] Figure 3 This diagram illustrates yet another client discovery method according to an embodiment of the present disclosure. For example... Figure 3 As shown, the method provided in this disclosure embodiment may include the following steps.

[0165] Figure 3 The example applies to longitudinal federated learning initiated by NWDAF, i.e., the VFL training process is initiated by NWDAF. Possible server and client combinations in this scenario are as follows:

[0166] Training phase:

[0167] VFL server: MTLF;

[0168] VFL client: one MTLF, or one AF, or multiple MTLFs, or multiple MTLFs and one AF, or multiple AFs, or multiple MTLFs and multiple AFs.

[0169] Reasoning stage:

[0170] VFL server: ANLF;

[0171] VFL client: one ANLF, or one AF, or multiple ANLFs, or multiple ANLFs and one AF, or multiple AFs, or multiple ANLFs and multiple AFs.

[0172] In this embodiment of the disclosure, the server and the client are not the same network element. For example, the MTLF acting as the server and the MTLF acting as the client are not the same MTLF.

[0173] In this embodiment of the disclosure, each option of the VFL client in the training phase corresponds one-to-one with each option of the VFL client in the inference phase. For example, if the training phase client is an MTLF, then the inference phase client is an ANLF. As another example, if the training phase client is an AF, then the inference phase client is the same AF. As another example, if the training phase client is multiple MTLFs, then the inference phase client is the multiple ANLFs corresponding to those multiple MTLFs. As another example, if the training phase client is multiple MTLFs and one AF, then the inference phase client is the multiple ANLFs corresponding to those multiple MTLFs and the same AF. As another example, if the training phase client is multiple AFs, then the inference phase client is those multiple AFs. As another example, if the training phase client is multiple MTLFs and multiple AFs, then the inference phase client is the multiple ANLFs corresponding to those multiple MTLFs and the multiple AFs.

[0174] In this embodiment of the disclosure, one MTLF corresponds to one ANLF, and one ANLF may correspond to multiple MTLFs.

[0175] In S31, the Consumer NF sends a request to AnLF (VFL Server for inference) carrying the Analytics ID.

[0176] In this embodiment of the disclosure, Consumer NF is a consumer network element, which may be at least one of core network elements such as PCF (Policy Control Function), AMF, and SMF (Session Management function).

[0177] Figure 3 In this embodiment, it is assumed that AnLF (Analysis Logic Function Network Element) acts as the inference vertical federated learning server, i.e., VFL Server for inference. The Consumer NF sends an Analysis Identifier Request to AnLF (VFL Server for inference), which carries the Analysis Identifier (Analytics ID). AnLF (VFL Server for inference) receives the Analysis Identifier Request.

[0178] In S32, AnLF (VFL Server for inference) believes that VFL should be used and discovers the server network element for VFL training.

[0179] In this embodiment of the disclosure, if AnLF (VFL Server for inference) deems it necessary to adopt vertical federated learning, then the network element discovery process is used to discover MTLF network elements (i.e., VFL Server for training (MTLF), the VFL Server in the training phase, i.e., at this time it is assumed that MTLF is the VFL Server in the training phase) from the NRF.

[0180] In this embodiment of the disclosure, the conditions under which AnLF (VFL Server for inference) deems vertical federated learning necessary can be determined based on internal configuration, and this disclosure does not impose any limitations on this.

[0181] In S33, AnLF (VFL Server for inference) sends a longitudinal federated learning model training request to VFL Server for training (MTLF).

[0182] In this embodiment of the disclosure, after AnLF (VFL Server for inference) discovers VFL Server for training (MTLF), it sends a request (which can be called a vertical federated learning model training request or model acquisition request) to MTLF, which is the VFL Server in the training phase, to request to acquire the vertical federated learning model or to request to train the vertical federated learning model.

[0183] In an exemplary embodiment, the model acquisition request may include at least one of Analytics ID, filter information, model ID, etc.

[0184] In S34, the VFL Server for training (MTLF) discovers other client network elements that are training VFL models.

[0185] In this embodiment of the disclosure, if there is no model that meets the requirements inside the VFL Server for training (MTLF) (i.e., no matching model is found according to the information carried in the training request of the above-mentioned longitudinal federated learning model), the network element discovery process is used to discover other network elements that can serve as clients in the training phase from the NRF, that is, to discover VFL client for training (MTLF or AF), and the discovered MTLF or AF can be used as VFL clients in the training phase.

[0186] In S35, the VFL Server for training (MTLF) and the discovered VFL client for training (MTLF or AF) train the VFL model.

[0187] In this embodiment of the disclosure, the VFL server (e.g., MTLF) and VFL client in the training phase complete the training process of the VFL model (i.e., the model of longitudinal federated learning).

[0188] For example, the VFL Server for training (MTLF) sends a request message to the VFL client for training (MTLF or AF), triggering the VFL client for training (MTLF or AF) to perform vertical federated learning to obtain the model. Specifically, this can involve VFL operations between the VFL Server for training (MTLF) and at least one VFL client for training (MTLF or AF), such as an interactive iterative process of vertical federated learning, to obtain the aforementioned model. This model is obtained through vertical federated learning; therefore, it can also be called a vertical federated learning model or a model of vertical federated learning.

[0189] In S36, the VFL Server for training (MTLF) returns the trained model and the NF IDs of the client network elements during the training phase to the AnLF (VFL Server for inference).

[0190] In this embodiment of the disclosure, after the VFL Server for training (MTLF) obtains the above model through vertical federated learning, the VFL Server for training (MTLF) sends model information to AnLF (VFL Server for inference), such as model file information, model file download address information, or model download address, etc.

[0191] In this embodiment of the disclosure, the NF ID(s) of the client network element during the training phase returned by the VFL Server for training (MTLF) to the AnLF (VFL Server for inference) includes at least one of the following:

[0192] (1) MTLF ID(s) participating in training;

[0193] (2) AF IDs participating in training.

[0194] In this embodiment of the disclosure, the above model can be a model used for communication service processing in a communication system, for example: it can be a model used for data analysis, or a model used for inference tasks, or a model used for channel estimation, or a model used for information prediction, etc.

[0195] In S37, AnLF (VFL Server for inference) uses the NF ID of the client network element in the training phase to discover the client in the inference phase.

[0196] In this embodiment of the disclosure, after receiving the NF ID(s) of the client network element in the training phase sent by the VFL Server for Training (MTLF), AnLF (VFL Server for Inference) can utilize... Figure 2 In the embodiment, steps S23 and S24 discover the client network element in the inference phase.

[0197] In S38, AnLF (VFL Server for inference) and VFL client for inference (AnLFor AF) implement VFL model inference.

[0198] In this embodiment of the disclosure, the server and client complete the inference process during the inference phase and obtain the inference result. For example, during the inference phase, the server and client input the collected information into the trained VFL model to obtain the inference result.

[0199] In S39, AnLF (VFL Server for inference) returns the analysis results to Consumer NF.

[0200] In some embodiments, AnLF (VFL Server for Inference) returns the inference result to the Consumer NF that sent the analysis identifier request. In other embodiments, AnLF (VFL Server for Inference) may further process the inference result (e.g., perform format processing to meet the fixed format requirements of analysis results) to obtain the analysis result, and then send the analysis result to the Consumer NF.

[0201] The method provided in this disclosure, on the one hand, enables model training function nodes to obtain information about the model for vertical federated learning, thereby improving the performance of network nodes in model training and solving the problem of being unable to obtain the model due to data privacy. For example, if a model training function node needs to provide a model for an AOI specified by a model inference function node, but the model training function node finds that it may not be able to obtain all or part of the training data in the AOI due to data silos, it triggers vertical federated learning to the VFL Server. On the other hand, by returning the NF ID of the VFL client in the training phase to the VFL Server in the inference phase through the VFL Server in the training phase, the VFL Server in the inference phase can more quickly and accurately discover the corresponding AnLF or AF as the VFL client in the inference phase.

[0202] Figure 4 A schematic diagram of another client discovery method in an embodiment of this disclosure is shown.

[0203] In this embodiment, the NWDAF can possess either data analysis capability or model training capability, or both. Under this logical functional decomposition requirement, NWDAFs supporting different capabilities can be considered as different logical functional modules, namely: a) AnLF. An NWDAF containing AnLF can perform data inference, generate analytical information (such as generating statistical analysis of the past or predictive information for the future based on the requests or subscriptions of analytical consumers), and provide the analytical information / analysis results / inference results as its services to analytical consumers. b) MTLF. An NWDAF containing MTLF can train an initial analytical model and provide the trained analytical model to AnLF. Analytical service consumers can discover NWDAFs supporting the required capabilities (AnLF capability, MTLF capability, or both) through the NRF based on the NF discovery and selection mechanism under the service-oriented architecture. Furthermore, the NWDAF will register its supported Analytics ID, service area / service scope, and other information as part of its NF Profile in the NRF to enable analytical service consumers to find a specific NWDAF. Figure 4 The assumption is that Consumer NF finds an AnLF.

[0204] Figure 4 The example applies to vertical federated learning initiated by AF, in which possible server and client combinations are as follows:

[0205] Training phase:

[0206] VFL server: AF;

[0207] VFL client: One or more MTLFs. When the server during the training phase is AF, the client is one or more MTLFs.

[0208] Reasoning stage:

[0209] VFL server: AF; when the server during the training phase is AF, the server during the inference phase is the same AF.

[0210] VFL client: One ANLF or multiple ANLFs.

[0211] like Figure 4 As shown, the method provided in this disclosure embodiment may include the following steps.

[0212] In S41, Consumer NF sends a request to AnLF, which carries the Analytics ID.

[0213] Understandable Figure 4 The AnLF that receives the analysis identifier request sent by the Consumer NF in S41 and the AnLF (VFL client for inference(AnLF)) that serves as the client in the inference phase are not the same AnLF.

[0214] In an exemplary embodiment, the Consumer NF also discovers the AnLF via the NRF and then sends an analysis identification request to the AnLF. Other details can be found in S31 above.

[0215] In S42, AnLF believes that VFL should be used and discovers the server network element for VFL training.

[0216] In this embodiment of the disclosure, after receiving the analysis identification request sent by the Consumer NF, if AnLF deems it necessary to employ vertical federated learning (based on the information provided in the analysis identification request in S41 and some internal configuration to determine whether it is necessary), and this vertical federated learning needs to be initiated by the AF, or as... Figure 3 As shown in step S32, AnLF uses the network element discovery process to discover from NRF that the network element that can serve as the server in the training phase is the AF network element.

[0217] In S43, AnLF sends a vertical federated learning model training request to AF (VFL Server for training and inference, i.e., AF acts as the VFL Server for both the training and inference phases).

[0218] In this embodiment of the disclosure, AnLF sends a vertical federated learning request or a vertical federated learning model training request to the AF (VFL Server for training and inference) to request analysis results or inference results.

[0219] In S44, AF (VFL Server for training and inference) discovers other client network elements that are training VFL models.

[0220] In this embodiment of the disclosure, after receiving a vertical federated learning request or a vertical federated learning model training request, if there are no analysis results or inference results that meet the requirements internally, the AF (VFL Server for training and inference) uses the network element discovery process to discover other network elements that can serve as clients in the training phase from the NRF, such as VFLclient for training (MTLF), that is, it is assumed that MTLF is discovered as a VFL client in the training phase.

[0221] In S45, the AF (VFL Server for training and inference) and the discovered VFL client for training (MTLF) perform VFL model training.

[0222] Figure 4 In this embodiment, it is assumed that AF initiates the VFL training process. Here, AF is the labeled active participant, while NWDAF is the passive participant.

[0223] In Virtual Frontier Function (VFL), active and passive participants are required to have the same samples and different features with the same sample labels before model training. This disclosure proposes a preparation process for negotiating between active participants (AF) and passive participants (NWDAF) to ensure they share the same sample space, such as the same UE, and to identify features before VFL model training. In this process, the NEF (Network Exposure Function) is responsible for discovering candidate passive participants and checking their willingness. During VFL model training, the AF and NWDAF exchange intermediate data and then compute losses and gradients to update their respective local models. The VFL model training process iterates until the AF determines that the model performance meets the requirements based on comparisons with the labels.

[0224] For other details regarding S45, please refer to the above. Figure 3S35 in the embodiment.

[0225] In S46, the AF (VFL Server for training and inference) uses the NF ID of the client network element in the training phase to discover the client in the inference phase.

[0226] In this embodiment of the disclosure, the AF (VFL Server for training and inference) utilizes the NF ID(s) of the VFL client during the training phase according to... Figure 2 In the embodiment, steps S23 and S24 discover the client in the inference phase. At this time, the NF ID(s) of the VFL client in the training phase includes the MTLF ID(s) that participated in the training.

[0227] In S47, AF (VFL Server for training and inference) and VFL client for training (MTLF) implement VFL model inference and obtain inference results.

[0228] For details, please refer to [link / reference]. Figure 3 S38 in the embodiment.

[0229] In S48, the AF (VFL Server for training and inference) returns the inference results to AnLF.

[0230] The AF (VFL Server for training and inference) sends the inference results to AnLF. AnLF receives the inference results.

[0231] In S49, AnLF returns the analysis results to Consumer NF.

[0232] AnLF further processes the received inference results to obtain analysis results, making the analysis results conform to a fixed format, and then sends the analysis results to Consumer NF.

[0233] The method provided in this disclosure allows the VFL Server in the training phase to also serve as the VFL Server in the inference phase. After determining the VFL client in the training phase, the method discovers the VFL client in the inference phase based on the NF ID of the VFL client in the training phase. This enables faster and more accurate discovery of the corresponding AnLF to serve as the VFL client in the inference phase.

[0234] Figure 5 This diagram illustrates yet another client discovery method according to an embodiment of the present disclosure. Figure 5 As shown, the method provided in this disclosure embodiment may include the following steps.

[0235] In some embodiments, if the ANLF does not include the corresponding MTLF ID(s) in its NF profile, and therefore the MTLF ID(s) corresponding to the ANLF is not registered in the NRF, that is, it is not registered. Figure 2 In step S21 of the embodiment, when the VFL server in the inference phase searches for the VFL client in the inference phase, it cannot find the accurate ANLF based on the MTLF ID(s) from the training phase to use for the VFL client in the inference phase. In this embodiment of the disclosure, to solve this problem, in... Figure 2 In steps S23 and S24 of the embodiment, the NRF returns candidate ANLF(s) (i.e., candidate analytical logic function network elements) based on other conditions. At this time, utilizing... Figure 5 The method provided in the embodiments further determines the ANLF from the candidate ANLF(s) to serve as the VFLclient for the inference phase. That is, for Figure 3 Step S37 and in the embodiment Figure 4 Step S46 in the embodiment is extended.

[0236] In this embodiment of the disclosure, other conditions may include, for example, the geographic information of the service provider during the search. The NRF can determine some candidate network elements, i.e., candidate ANLF(s), based on this information.

[0237] In S51, the VFL Server for inference (AF or ANLF, i.e., AF or ANLF as the VFL Server in the inference phase) discovers network elements that can act as clients from the NRF (i.e., candidate analytical logic function network elements that may be able to act as VFL clients in the inference phase).

[0238] In this embodiment of the disclosure, the VFL server (inference phase) utilizes the network element discovery process to discover potential network elements (i.e., candidate ANLF(s) that may serve as VFL clients in the inference phase) from the NRF. Since the NF profile of the ANLF does not contain information about its associated MTLF, the NRF returns candidate client network elements (i.e., candidate analysis logic function network elements) based on other conditions, hereinafter referred to as VFL client for inference (Candidate).

[0239] In S52, the VFL Server for inference (AF or ANLF) sends a VFL inference request to the VFL client for inference (Candidate), which includes the MTLF ID(s).

[0240] The VFL Server for inference (AF or ANLF) sends a VFL request (i.e., a VFL inference request) to the candidate network element (i.e., the VFL client for inference (Candidate)). This request includes the MTLF ID(s) from the training phase. Specifically, the longitudinal federated learning inference request includes the identifier of the training longitudinal federated learning client network element, which in turn includes the identifier of the model training logic function. The VFL client for inference (Candidate) receives this VFL inference request to obtain the MTLF ID(s) from the training phase.

[0241] In S53, the VFL client for inference (Candidate) uses the MTLF ID to determine whether to participate in VFL inference.

[0242] In this embodiment of the disclosure, the candidate VFL client network element (i.e., VFL client for inference (Candidate)) determines whether it participates in VFL inference based on the MTLF ID(s) and other information during the training phase.

[0243] In an exemplary embodiment, the candidate VFL client network element determines whether it participates in VFL inference by including at least one of the following:

[0244] (1) The MTLF ID(s) during the training phase are compared sequentially with the MTLF ID(s) configured within the candidate VFL client network element to determine whether they match. If they match, the candidate VFL client network element is determined to participate in VFL inference; if they do not match, the candidate VFL client network element is determined not to participate in VFL inference.

[0245] (2) The candidate VFL client network element further utilizes the network element discovery process from the NRF to discover MTLFs that can provide services to it. Then, it compares the MTLF ID(s) from the training phase with the MTLF ID(s) returned by the NRF that can provide services to it to determine whether they match. If they match, the candidate VFL client network element is determined to participate in VFL inference; if they do not match, the candidate VFL client network element is determined not to participate in VFL inference.

[0246] In S54, the VFL client for inference (Candidate) returns to the VFL Server for inference (AF or ANLF) whether it has joined the VFL inference process.

[0247] The VFL client for inference (Candidate) returns response information to the VFL server for inference (AF or ANLF). This response information includes indication information indicating whether the client participates in VFL inference, i.e., whether it joins the VFL inference process.

[0248] After receiving the response information, the VFL Server for Inference (AF or ANLF) selects the VFL client for inference (Candidate) that it is instructed to join the VFL inference process as the VFL client for the inference phase, based on the instructions in the response information. Once the VFL client for the inference phase is determined, the other steps can be referred to the above embodiment and will not be repeated here.

[0249] The method provided in this disclosure allows the ANLF to send its internally configured MTLF ID(s) to the VFL server during the training phase when registering its NF profile with the NRF. This enables the NRF to find matching VFL client network elements based on the MTLF ID, and thus allows the VFL server during the inference phase to discover client network elements based on the client NF ID. If the ANLF has not registered its pre-configured MTLF ID with the NRF, the VFL server sends the MTLF ID(s) from the training phase when requesting inference from a potential VFL client. The potential VFL client then uses this MTLF ID(s) to determine whether it should participate in the longitudinal federated learning inference process.

[0250] Figure 6 A flowchart illustrating another client discovery method in an embodiment of this disclosure is shown. Figure 6 The method provided in the embodiments can be executed by training a longitudinal federated learning server. For example... Figure 6 As shown, the method provided in this embodiment includes the following steps.

[0251] In S610, the inference vertical federated learning server obtains the network element identifier of the training vertical federated learning client.

[0252] In an exemplary embodiment, the training longitudinal federated learning server is a model training logic function network element. In an exemplary embodiment, the training longitudinal federated learning client is one of the following: a model training logic function network element, or an application function network element, or multiple model training logic function network elements, or multiple model training logic function network elements and one application function network element, or multiple application function network elements, or multiple model training logic function network elements and multiple application function network elements.

[0253] In an exemplary embodiment, the inference vertical federated learning server is an analysis logic function network element. In an exemplary embodiment, the inference vertical federated learning client network element is one of the following: one analysis logic function network element, or one application function network element, or multiple analysis logic function network elements, or multiple analysis logic function network elements and one application function network element, or multiple application function network elements, or multiple analysis logic function network elements and multiple application function network elements.

[0254] In an exemplary embodiment, the method provided in this disclosure further includes: a model training logic function network element serving as a training vertical federated learning server, receiving a model acquisition request sent by an analysis logic function network element serving as an inference vertical federated learning server; determining a training vertical federated learning client network element from a network storage function network element; and the model training logic function network element serving as a training vertical federated learning server and the training vertical federated learning client network element completing model training.

[0255] Figure 6 Other aspects of the embodiments can be found in the other embodiments described above, and will not be repeated here.

[0256] Figure 7 A flowchart of yet another client discovery method according to an embodiment of this disclosure is shown. Figure 7 The method provided in this embodiment can be executed by a network storage function network element. For example... Figure 7 As shown, the method provided in this disclosure embodiment may include the following steps.

[0257] In S710, a network element discovery request sent by the inference longitudinal federated learning server is received, the network element discovery request including the network element identifier of the training longitudinal federated learning client.

[0258] In S720, the matching inference longitudinal federated learning client network element identifier is searched based on the network element discovery request.

[0259] In S730, the matching inference vertical federated learning client element identifier is sent to the inference vertical federated learning server.

[0260] In an exemplary embodiment, the method provided in this disclosure further includes: receiving a registration request from an analysis logic function network element. The registration request includes network element configuration of the analysis logic function network element, and the network element configuration includes at least one of the following: network data analysis function network element type, analysis identifier, network data analysis function address information, service range, and vertical federated learning capability information.

[0261] In an exemplary embodiment, the vertical federated learning capability information includes at least one of the following: vertical federated learning inference indication information, which indicates that the analysis logic function network element supports vertical federated learning inference function and can participate in the vertical federated inference process; vertical federated learning capability indication information, which indicates that the analysis logic function network element supports vertical federated learning function and can participate in the vertical federated learning process; vertical federated learning server capability indication information, which indicates that the analysis logic function network element can participate in the vertical federated learning process as a vertical federated learning server; vertical federated learning client capability indication information, which indicates that the analysis logic function network element can participate in the vertical federated learning process as a vertical federated learning client; a corresponding model training logic function identifier, which indicates the model training logic function identifier configured inside the analysis logic function network element (wherein, when the ANLF needs to interact with the MTLF, it will send a request to the corresponding MTLF according to the MTLF ID(s); or, indicating the model training logic function identifier configured inside the analysis logic function network element that can be used for vertical federated learning.

[0262] In an exemplary embodiment, the network element discovery request includes at least one of the following: vertical federated learning inference indication information, which indicates that the inference vertical federated learning client network element corresponding to the inference vertical federated learning client network element identifier needs to support the vertical federated learning inference function and be able to participate in the vertical federated inference process; vertical federated learning capability indication information, which indicates that the inference vertical federated learning client network element corresponding to the inference vertical federated learning client network element identifier needs to support the vertical federated learning function and be able to participate in the vertical federated learning process; vertical federated learning client capability indication information, which indicates that the inference vertical federated learning client network element corresponding to the inference vertical federated learning client network element identifier needs to be able to participate in the vertical federated learning process as a vertical federated learning client; model training logic function identifier, which indicates that the model training logic function network element corresponding to the inference vertical federated learning client network element identifier needs to include the model training logic function network element specified by the model training logic function identifier, wherein the model training logic function network element identifier includes the model training logic function identifier; network data analysis function network element type; analysis identifier; service scope.

[0263] In an exemplary embodiment, finding a matching inference vertical federated learning client network element identifier based on the network element discovery request includes at least one of the following: determining that the vertical federated learning inference indication information in the network element discovery request matches the information in the network element configuration; determining that the vertical federated learning capability indication information in the network element discovery request matches the information in the network element configuration; determining that the vertical federated learning client capability indication information in the network element discovery request matches the information in the network element configuration; determining that the model training logic function identifier in the network element discovery request matches the information in the network element configuration; determining that the network data analysis function network element type in the network element discovery request matches the information in the network element configuration; determining that the analysis identifier in the network element discovery request matches the information in the network element configuration; and determining that the service scope in the network element discovery request matches the information in the network element configuration.

[0264] Figure 7 Other aspects of the embodiments can be found in the other embodiments described above, and will not be repeated here.

[0265] Figure 8 A flowchart of yet another client discovery method according to an embodiment of this disclosure is shown. Figure 8 The method provided in the embodiments can be executed by analysis logic function network elements. For example... Figure 8 As shown, the method provided in this disclosure embodiment may include the following steps.

[0266] In S810, a registration request is sent to the network storage function network element. The registration request includes the network element configuration of the analysis logic function network element. The network element configuration includes at least one of the following: network data analysis function network element type, analysis identifier, network data analysis function address information, service range, and vertical federated learning capability information. The vertical federated learning capability information includes a corresponding model training logic function identifier, which indicates the model training logic function identifier configured internally within the analysis logic function network element; or, it indicates the model training logic function identifier configured internally within the analysis logic function network element that can be used for vertical federated learning.

[0267] Figure 8 Other aspects of the embodiments can be found in the other embodiments described above, and will not be repeated here.

[0268] Figure 9 A schematic diagram of a client discovery device according to an embodiment of this disclosure is shown. Figure 9 The client discovery device 900 provided in this embodiment can be applied to a reasoning-based longitudinal federated learning server. For example... Figure 9As shown, the client discovery device 900 may include an acquisition unit 910 and a processing unit 920. The acquisition unit 910 is used to acquire the training longitudinal federated learning client element identifier from the training longitudinal federated learning server. The processing unit 920 is used to discover matching inference longitudinal federated learning client elements based on the training longitudinal federated learning client element identifier.

[0269] Figure 9 Other aspects of the embodiments can be found in the other embodiments described above, and will not be repeated here.

[0270] Figure 10 A schematic diagram of another client discovery device is shown in an embodiment of this disclosure. Figure 10 The client discovery device 1000 provided in this embodiment can be applied to training a longitudinal federated learning server. For example... Figure 10 As shown, the client discovery device 1000 may include a processing unit 1010. The processing unit 1010 is used to enable the inference longitudinal federated learning server to obtain the network element identifier of the training longitudinal federated learning client.

[0271] Figure 10 Other aspects of the embodiments can be found in the other embodiments described above, and will not be repeated here.

[0272] Figure 11 A schematic diagram of yet another client discovery device is shown in an embodiment of this disclosure. Figure 11 The client discovery device 1100 provided in this embodiment can be applied to network storage function network elements. For example... Figure 11 As shown, the client discovery device 1100 may include a receiving unit 1110, a processing unit 1120, and a sending unit 1130. The receiving unit 1110 receives a network element discovery request sent by the inference longitudinal federated learning server, the network element discovery request including a network element identifier for the training longitudinal federated learning client. The processing unit 1120 searches for a matching inference longitudinal federated learning client network element identifier according to the network element discovery request. The sending unit 1130 sends the matching inference longitudinal federated learning client network element identifier to the inference longitudinal federated learning server.

[0273] Figure 11 Other aspects of the embodiments can be found in the other embodiments described above, and will not be repeated here.

[0274] Figure 12 A schematic diagram of another client discovery device is shown in an embodiment of this disclosure. Figure 12 The client discovery device 1200 provided in this embodiment can be applied to analyze logical function network elements. For example... Figure 12As shown, the client discovery device 1200 may include a sending unit 1210. The sending unit 1210 is used to send a registration request to a network storage function network element. The registration request includes the network element configuration of the analysis logic function network element. The network element configuration includes at least one of the following: network data analysis function network element type, analysis identifier, network data analysis function address information, service range, and vertical federated learning capability information. The vertical federated learning capability information includes a corresponding model training logic function identifier, which indicates the model training logic function identifier configured internally by the analysis logic function network element; or, indicates a model training logic function identifier configured internally by the analysis logic function network element that can be used for vertical federated learning.

[0275] Figure 12 Other aspects of the embodiments can be found in the other embodiments described above, and will not be repeated here.

[0276] This disclosure also provides a communication device, including: a processor; and a memory for storing executable instructions of the processor. The processor is configured to execute the methods described in any embodiment of this disclosure by executing the executable instructions.

[0277] The following reference Figure 13 This describes a communication device 1000 according to such an embodiment of the present disclosure. Figure 13 The communication device 1000 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments disclosed herein.

[0278] like Figure 13 As shown, the communication device 1000 is presented in the form of a general-purpose computing device. The components of the communication device 1000 may include, but are not limited to: at least one processing unit 1010, at least one storage unit 1020, and a bus 1030 connecting different system components (including storage unit 1020 and processing unit 1010).

[0279] The storage unit 1020 stores program code that can be executed by the processing unit 1010, causing the processing unit 1010 to perform the steps described in the "Exemplary Methods" section above according to various exemplary embodiments of this disclosure.

[0280] Storage unit 1020 may include readable media in the form of volatile storage units, such as random access memory (RAM) 10201 and / or cache memory 10202, and may further include read-only memory (ROM) 10203.

[0281] Storage unit 1020 may also include a program / utility 10204 having a set (at least one) program module 10205, such program module 10205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0282] Bus 1030 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the multiple bus structures.

[0283] The communication device 1000 can also communicate with one or more external devices 1040 (e.g., keyboard, pointing device, Bluetooth device, etc.), one or more devices that enable a user to interact with the communication device 1000, and / or any device that enables the communication device 1000 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via the input / output (I / O) interface 1050. Furthermore, the communication device 1000 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via the network adapter 1060. As shown, the network adapter 1060 communicates with other modules of the communication device 1000 via the bus 1030.

[0284] This disclosure also provides a computer program product including a computer program that, when run, executes the methods described in any embodiment of this disclosure. This disclosure further provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the methods described in any embodiment of this disclosure.

Claims

1. A client discovery method, characterized in that, The method is executed by the inference longitudinal federated learning server, and the method includes: Obtain the network element identifier of the training vertical federated learning client from the training vertical federated learning server; Based on the training longitudinal federated learning client element identifier, a matching inference longitudinal federated learning client element is discovered.

2. The method according to claim 1, characterized in that, The training longitudinal federated learning client element identifier includes at least one of the following: Model training logic function identifier; Application function identifier.

3. The method according to claim 1, characterized in that, Based on the training longitudinal federated learning client element identifier, matching inference longitudinal federated learning client elements are discovered, including: Send a network element discovery request to the network storage function network element, the network element discovery request including the network element identifier of the training longitudinal federated learning client; The network storage function network element receives the inference vertical federated learning client network element identifier returned in response to the network element discovery request, and the inference vertical federated learning client network element identifier corresponds to the inference vertical federated learning client network element.

4. The method according to claim 3, characterized in that, The network element discovery request includes at least one of the following: Vertical federated learning inference instruction information, which is used to indicate that the inference vertical federated learning client network element needs to support the vertical federated learning inference function and be able to participate in the vertical federated learning inference process; Vertical federated learning capability indication information, which is used to indicate that the inference vertical federated learning client network element needs to support the vertical federated learning function and be able to participate in the vertical federated learning process; Vertical federated learning client capability indication information, which is used to indicate that the inference vertical federated learning client network element needs to be able to participate in the vertical federated learning process as a vertical federated learning client. The model training logic function identifier is used to indicate that the model training logic function network element corresponding to the inference longitudinal federated learning client network element contains the model training logic function network element specified by the training longitudinal federated learning client network element identifier. Network data analysis function network element types; Analysis identifiers; Service scope.

5. The method according to claim 1, characterized in that, The training vertical federated learning server is a model training logic function network element; The training vertical federated learning client can be one of the following: a model training logic function network element, or an application function network element, or multiple model training logic function network elements, or multiple model training logic function network elements and one application function network element, or multiple application function network elements, or multiple model training logic function network elements and multiple application function network elements.

6. The method according to claim 5, characterized in that, The inference vertical federated learning server is an analysis logic function network element; The inference vertical federated learning client network element can be one of the following: an analysis logic function network element, or an application function network element, or multiple analysis logic function network elements, or multiple analysis logic function network elements and one application function network element, or multiple application function network elements, or multiple analysis logic function network elements and multiple application function network elements.

7. The method according to claim 6, characterized in that, Also includes: The analysis logic function network element, which serves as the inference vertical federated learning server, receives the analysis identifier request sent by the consumer network element. In response to the analysis identifier request, the network storage function network element discovers the model training logic function network element that can serve as the training server for longitudinal federated learning; Send a model retrieval request to the model training logic function network element that serves as the training server for vertical federated learning, in order to request the model for vertical federated learning.

8. The method according to claim 7, characterized in that, Also includes: It receives information about the trained model from the model training logic function network element, which serves as the training server for vertical federated learning.

9. The method according to claim 7, characterized in that, Also includes: The inference vertical federated learning server and the inference vertical federated learning client network element complete the inference process and obtain the inference result.

10. The method according to claim 1, characterized in that, The training vertical federated learning server is an application function network element; The training vertical federated learning client can be one of the following: a model training logical function network element, or multiple model training logical function network elements.

11. The method according to claim 10, characterized in that, The inference vertical federated learning server is an application function network element; The inference vertical federated learning client network element is one of the following: one analysis logic function network element, or multiple analysis logic function network elements.

12. The method according to claim 11, characterized in that, Also includes: The application function network element receives the vertical federated learning request sent by the analysis logic function network element; In response to the vertical federated learning request, the training vertical federated learning client is discovered from the network storage function element; The application function network element and the training longitudinal federated learning client are used to perform longitudinal federated learning training to obtain the trained model; The application function network element uses the training longitudinal federated learning client identifier of the training longitudinal federated learning client to discover the matching inference longitudinal federated learning client network element.

13. The method according to claim 12, characterized in that, The network element identifiers for the training longitudinal federated learning client include: Identifiers of the training logic functions of the model involved in training.

14. The method according to claim 12, characterized in that, Also includes: The inference process is completed through the application function network element and the inference vertical federated learning client network element to obtain the inference result; The inference result is returned to the analysis logic function element that sent the vertical federated learning request.

15. The method according to claim 1, characterized in that, Based on the training longitudinal federated learning client element identifier, matching inference longitudinal federated learning client elements are discovered, including: Send a vertical federated learning inference request to the candidate analysis logic function network element; Receive the response information returned by the candidate analysis logic function network element in response to the vertical federated learning inference request; The inference vertical federated learning client network element is determined from the candidate analysis logic function network elements based on the response information.

16. The method according to claim 15, characterized in that, Also includes: The candidate analysis logic function network element is determined from the network storage function network element.

17. The method according to claim 15, characterized in that, The vertical federated learning inference request includes the training vertical federated learning client element identifier, which includes the model training logic function identifier. The response information includes indications as to whether the corresponding candidate analysis logic function element is included in the vertical federated learning inference process; The candidate analysis logic function network element is used to determine whether to participate in the longitudinal federated learning inference based on the network element identifier of the training longitudinal federated learning client, so as to generate the indication information.

18. The method according to any one of claims 15 to 17, characterized in that, The candidate analysis logic function element is used to determine whether to participate in longitudinal federated learning inference based on at least one of the following: The identifier of the training longitudinal federated learning client network element is compared with the identifier of the model training logic function configured inside the candidate analysis logic function network element in turn to determine whether they match. The network storage function element discovers model training logic function elements that can provide services to the candidate analysis logic function elements. The identifier of the training longitudinal federated learning client element and the identifier of the model training logic function returned by the network storage function element are compared in turn to determine whether they match.

19. The method according to claim 1, characterized in that, Based on the training longitudinal federated learning client element identifier, matching inference longitudinal federated learning client elements are discovered, including: When the training longitudinal federated learning client element identifier is an application function identifier, the application function element corresponding to the application function identifier is determined to be the inference longitudinal federated learning client element.

20. A client discovery method, characterized in that, The method is executed by a training longitudinal federated learning server, and the method includes: This enables the inference vertical federated learning server to obtain the network element identifier of the training vertical federated learning client.

21. The method according to claim 20, characterized in that, The training vertical federated learning server is a model training logic function network element; The training longitudinal federated learning client can be one of the following: a model training logic function network element, or an application function network element, or multiple model training logic function network elements, or multiple model training logic function network elements and one application function network element, or multiple application function network elements, or multiple model training logic function network elements and multiple application function network elements.

22. The method according to claim 21, characterized in that, The inference vertical federated learning server is an analysis logic function network element; The inference vertical federated learning client network element can be one of the following: an analysis logic function network element, or an application function network element, or multiple analysis logic function network elements, or multiple analysis logic function network elements and one application function network element, or multiple application function network elements, or multiple analysis logic function network elements and multiple application function network elements.

23. The method according to claim 22, characterized in that, Also includes: As a model training logic function network element that serves as the training vertical federated learning server, it receives a model acquisition request sent by the analysis logic function network element that serves as the inference vertical federated learning server. If there is no model that satisfies the model acquisition request internally, the training vertical federated learning client network element is determined from the network storage function network element; The model training logic function network element, which serves as the server-side network element for training vertical federated learning, and the client-side network element for training vertical federated learning complete the model training.

24. A client discovery method, characterized in that, The method is executed by a network storage function network element, and the method includes: Receive network element discovery requests sent by the inference longitudinal federated learning server, wherein the network element discovery requests include network element identifiers for the training longitudinal federated learning client; The network element discovery request is used to find the matching inference vertical federated learning client network element identifier; The matching inference vertical federated learning client element identifier is sent to the inference vertical federated learning server.

25. The method according to claim 24, characterized in that, Also includes: Receive registration requests from the network element with analytical logic function; The registration request includes the network element configuration of the analysis logic function network element, and the network element configuration includes at least one of the following: network data analysis function network element type, analysis identifier, network data analysis function address information, service range, and vertical federated learning capability information.

26. The method according to claim 25, characterized in that, Vertical federated learning capability information includes at least one of the following: Vertical federated learning inference instruction information, which is used to indicate that the analysis logic function network element supports the vertical federated learning inference function and can participate in the vertical federated inference process; Vertical federated learning capability indication information, which is used to indicate that the analysis logic function network element supports the vertical federated learning function and can participate in the vertical federated learning process; Vertical federated learning server capability indication information, which is used to indicate that the analysis logic function network element can participate in the vertical federated learning process as a vertical federated learning server. Vertical federated learning client capability indication information, which is used to indicate that the analysis logic function network element can participate in the vertical federated learning process as a vertical federated learning client; The corresponding model training logic function identifier is used to indicate the model training logic function identifier configured inside the analysis logic function network element; or, it indicates the model training logic function identifier configured inside the analysis logic function network element that can be used for vertical federated learning.

27. The method according to claim 25 or 26, characterized in that, The network element discovery request includes at least one of the following: Vertical federated learning inference instruction information is used to indicate that the inference vertical federated learning client network element corresponding to the inference vertical federated learning client network element identifier needs to support the vertical federated learning inference function and be able to participate in the vertical federated inference process. Vertical federated learning capability indication information, which is used to indicate that the inference vertical federated learning client network element corresponding to the inference vertical federated learning client network element identifier needs to support the vertical federated learning function and be able to participate in the vertical federated learning process. Vertical federated learning client capability indication information, which is used to indicate that the inference vertical federated learning client network element corresponding to the inference vertical federated learning client network element identifier needs to be able to participate in the vertical federated learning process as a vertical federated learning client. The model training logic function identifier is used to indicate that the model training logic function network element corresponding to the inference longitudinal federated learning client network element identifier needs to include the model training logic function network element specified by the model training logic function identifier, and the training longitudinal federated learning client network element identifier includes the model training logic function identifier. Network data analysis function network element types; Analysis identifiers; Service scope.

28. The method according to claim 27, characterized in that, The network element identifier for the matching inference longitudinal federated learning client network element found based on the network element discovery request includes at least one of the following: Determine that the vertical federated learning inference instruction information in the network element discovery request matches the information in the network element configuration; Determine that the vertical federated learning capability indication information in the network element discovery request matches the information in the network element configuration; Determine that the vertical federated learning client capability indication information in the network element discovery request matches the information in the network element configuration; Determine that the model training logic function identifier in the network element discovery request matches the information in the network element configuration; Determine that the network data analysis function network element type in the network element discovery request matches the information in the network element configuration; Determine that the analysis identifier in the network element discovery request matches the information in the network element configuration; The service scope in the network element discovery request is determined to match the information in the network element configuration.

29. A client discovery method, characterized in that, The method is executed by an analysis logic function network element, and the method includes: Send a registration request to the network storage function element; The registration request includes the network element configuration of the analysis logic function network element, and the network element configuration includes at least one of the following: network data analysis function network element type, analysis identifier, network data analysis function address information, service range, and vertical federated learning capability information. The vertical federated learning capability information includes a corresponding model training logic function identifier, which is used to indicate the model training logic function identifier configured inside the analysis logic function network element; or, to indicate the model training logic function identifier configured inside the analysis logic function network element that can be used for vertical federated learning.

30. A client discovery device, characterized in that, The device is used in a longitudinal federated learning server for inference, and the device includes: The acquisition unit is used to obtain the network element identifier of the training longitudinal federated learning client from the training longitudinal federated learning server. The processing unit is used to discover matching inference longitudinal federated learning client elements based on the training longitudinal federated learning client element identifier.

31. A client discovery device, characterized in that, The device is used to train a longitudinal federated learning server, and the device includes: The processing unit is used to enable the inference longitudinal federated learning server to obtain the network element identifier of the training longitudinal federated learning client.

32. A client discovery device, characterized in that, The device is applied to a network storage function network element, and the device includes: The receiving unit is used to receive a network element discovery request sent by the inference longitudinal federated learning server, wherein the network element discovery request includes the network element identifier of the training longitudinal federated learning client. The processing unit is used to find the matching inference longitudinal federated learning client network element identifier according to the network element discovery request; The sending unit is used to send the matching inference vertical federated learning client element identifier to the inference vertical federated learning server.

33. A client discovery device, characterized in that, The device is used to analyze logical function network elements, and the device includes: The sending unit is used to send registration requests to network storage function network elements; The registration request includes the network element configuration of the analysis logic function network element, and the network element configuration includes at least one of the following: network data analysis function network element type, analysis identifier, network data analysis function address information, service range, and vertical federated learning capability information. The vertical federated learning capability information includes a corresponding model training logic function identifier, which is used to indicate the model training logic function identifier configured inside the analysis logic function network element; or, to indicate the model training logic function identifier configured inside the analysis logic function network element that can be used for vertical federated learning.

34. A communication device, characterized in that, include: processor; as well as Memory for storing the executable instructions of the processor; The processor is configured to execute the method of any one of claims 1 to 19, or the method of claims 20 to 23, or the method of claims 24 to 28, or the method of claim 29, by executing the executable instructions.

35. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method described in any one of claims 1 to 19, or implements the method described in claims 20 to 23, or implements the method described in claims 24 to 28, or implements the method described in claim 29.

36. A computer program product comprising a computer program that, when run, performs the method of any one of claims 1 to 19, or performs the method of claims 20 to 23, or performs the method of claims 24 to 28, or performs the method of claim 29.

Images