Machine learning model processing method and apparatus, and storage medium

Abstract

The application provides a machine learning model processing method and device and a storage medium. A UE creates a local machine learning model of a target application in advance according to a global machine learning model provided by a first network function entity, determines local training data related to the target application, trains the local machine learning model according to the local training data, and sends local model parameters of the trained local machine learning model to the first network function entity to update the global machine learning model by the first network function entity. Federated learning between the UE and the first network function entity for providing the model can be implemented, the performance of sharing, transmitting and training the machine learning model between the UE and the network is improved, and the rapidly developing communication service and application requirements are met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication, and particularly relates to a machine learning model processing method and device and a storage medium. BACKGROUND

[0002] In order to realize intelligent data analysis and control of a mobile communication network, a network data analysis function (NWDAF) is introduced in a 5G network. The NWDAF provides network data analysis services for other network function entities based on an artificial intelligence (AI) algorithm through interaction with other network function entities.

[0003] The NWDAF can perform federated learning with other NWDAFs, where federated learning is a machine learning framework that can effectively help multiple NWDAFs to perform data use and machine learning modeling under the requirements of user privacy protection and data security, so as to obtain an optimized machine learning (ML) model. Existing federated learning methods involving NWDAFs can only be implemented between NWDAF entities and are not suitable for rapidly developing communication business and application requirements. SUMMARY

[0004] The present application provides a machine learning model processing method, device and storage medium to realize federated learning between a UE and a first network function entity for providing a model.

[0005] In a first aspect, the present application provides a machine learning model processing method applied to a user equipment (UE), and the method comprises the following steps.

[0006] Determining local training data related to a target application;

[0007] Training a local machine learning model of the target application according to the local training data to obtain local model parameters of the trained local machine learning model, where the local machine learning model is obtained according to a global machine learning model trained by a first network function entity;

[0008] Sending the local model parameters to the first network function entity, where the local model parameters are used to update the global machine learning model.

[0009] Optionally, the method further comprises the following steps.

[0010] Obtaining global model parameters of an updated global machine learning model from the first network function entity, where the global model parameters are used to update the trained local machine learning model.

[0011] The global model parameter is obtained by the first network function entity updating the global machine learning model using at least one local model parameter sent by a UE.

[0012] Optionally, the sending of the local model parameter to the first network function entity comprises:

[0013] sending, by an application layer, model update information carrying the local model parameter to an application function (AF) entity, so as to send the model update information to the first network function entity by the AF entity; or

[0014] sending, by an application layer, model update information carrying the local model parameter to an application function (AF) entity, so as to send the model update information to the first network function entity by the AF entity through a network exposure function (NEF) entity; or

[0015] sending, by a non-access stratum layer, model update information carrying the local model parameter to an access and mobility management function (AMF) entity, so as to send the model update information to the first network function entity by the AMF entity.

[0016] Optionally, the model update information further comprises any one or more of the following: a UE identifier, an application identifier, and a first network function entity identifier, for use by the AF entity or the AMF entity to determine the first network function entity and / or the global machine learning model.

[0017] Optionally, the obtaining of the global model parameter of the updated global machine learning model from the first network function entity comprises:

[0018] receiving a model update response sent by the AF entity or the AMF entity, wherein the model update response comprises information of the global model parameter of the updated global machine learning model.

[0019] Optionally, the local machine learning model is obtained according to a global machine learning model trained by the first network function entity, comprising:

[0020] obtaining a first model file of the global machine learning model from the first network function entity;

[0021] creating a local machine learning model according to the first model file.

[0022] Optionally, the obtaining of the first model file of the global machine learning model from the first network function entity comprises:

[0023] sending, to the first network function entity, model obtaining information carrying a UE identifier and / or model description information;

[0024] obtaining information of a first model file of the global machine learning model from the first network function entity based on the model obtaining information.

[0025] Optionally, the sending, to the first network function entity, of the model obtaining information carrying the UE identifier and / or the model description information comprises:

[0026] sending, to an AF entity, the model obtaining information carrying the UE identifier and / or the model description information through an application layer, so as to send, by the AF entity, the model obtaining information to the first network function entity according to the UE identifier and / or the model description information; or

[0027] sending, to an AF entity, the model obtaining information carrying the UE identifier and / or the model description information through an application layer, so as to send, by the AF entity, the model obtaining information to the first network function entity according to the UE identifier and / or the model description information through a NEF entity; or

[0028] sending, to an AMF entity, the model obtaining information carrying the UE identifier and / or the model description information through a non-access stratum layer, so as to send, by the AMF entity, the model obtaining information to the first network function entity according to the UE identifier and / or the model description information.

[0029] Optionally, the obtaining of the first model file of the global machine learning model from the first network function entity comprises:

[0030] receiving a model obtaining response sent by the AF entity or the AMF entity, wherein the model obtaining response comprises information of the first model file, and the information of the first model file is used to obtain the first model file.

[0031] Optionally, the model description information comprises at least one of the following:

[0032] an application identifier, an application sub-characteristic identifier, time information, location information, and other model characteristic information.

[0033] In a second aspect, the present application provides a machine learning model processing method, applied to an AF entity, and the method comprises:

[0034] receiving model update information carrying local model parameters sent by a UE through an application layer, wherein the local machine learning model is a global machine learning model provided by a first network function entity to the UE in advance;

[0035] sending the model update information to the first network function entity, and the local model parameters are used to update the global machine learning model.

[0036] Optionally, the method further comprises:

[0037] obtaining, from the first network function entity, information of global model parameters of the updated global machine learning model, the global model parameters being used to update the trained local machine learning model;

[0038] sending a model update response to the UE through an application layer, wherein the model update response comprises the information of the global model parameters;

[0039] The global model parameters are obtained by the first network function entity by updating the global machine learning model using at least one local model parameter sent by a UE.

[0040] Optionally, the sending of the model update information to the first network function entity comprises:

[0041] If the AF entity is a trusted AF entity, the model update information is directly sent to the first network function entity; or

[0042] If the AF entity is an untrusted AF entity, the model update information is sent to the first network function entity through an NEF entity.

[0043] Optionally, the obtaining, from the first network function entity, information of global model parameters of the updated global machine learning model comprises:

[0044] The information of the global model parameters of the updated global machine learning model is directly obtained from the first network function entity or obtained from the first network function entity through an NEF entity.

[0045] Optionally, before the receiving of the model update information carrying the local model parameters sent by the UE through an application layer, the method further comprises:

[0046] receiving model obtaining information carrying UE identification and / or model description information sent by the UE through an application layer;

[0047] sending the model obtaining information to the first network function entity according to the UE identification and / or model description information;

[0048] obtaining, from the first network function entity, information of a first model file of the global machine learning model and sending the information to the UE through an application layer.

[0049] Optionally, the sending of the model obtaining information to the first network function entity according to the UE identification and / or model description information comprises:

[0050] If the AF entity is a trusted AF entity, a first network function entity serving the UE and / or a first network function entity capable of providing a global machine learning model satisfying the model description information is selected, and the model obtaining information is sent to the first network function entity; or

[0051] If the AF entity is an untrusted AF entity, the model obtaining information is sent to a NEF entity, and a first network function entity serving the UE and / or a first network function entity capable of providing a global machine learning model satisfying the model description information is selected through the NEF entity, and the model obtaining information is sent to the first network function entity.

[0052] Optionally, the information of the first model file of the global machine learning model obtained from the first network function entity and sent to the UE through an application layer comprises:

[0053] The information of the first model file of the global machine learning model is obtained from the first network function entity directly or through the NEF entity;

[0054] The model obtaining response is sent to the UE through the application layer, wherein the model obtaining response comprises the information of the first model file, and the information of the first model file is used to obtain the first model file.

[0055] In a third aspect, the present application provides a machine learning model processing method, applied to a first network function entity for providing a model, and the method comprises:

[0056] Obtaining local model parameters of a trained local machine learning model sent by a user equipment (UE), wherein the local machine learning model is obtained according to a global machine learning model trained by the first network function entity;

[0057] Updating the global machine learning model according to the local model parameters.

[0058] Optionally, after updating the global machine learning model according to the local model parameters, the method further comprises:

[0059] Sending global model parameters of the updated global machine learning model to the UE, wherein the global model parameters are used to update the trained local machine learning model.

[0060] The global model parameters are obtained by updating the global machine learning model by the first network function entity using at least one local model parameter sent by the UE.

[0061] Optionally, before obtaining the local model parameters of the trained local machine learning model sent by the UE, the method further comprises:

[0062] sending information of a first model file of the global machine learning model to the UE to create a local machine learning model according to the first model file by the UE.

[0063] In a fourth aspect, the present application provides a user equipment, comprising a memory, a transceiver and a processor:

[0064] a memory for storing a computer program; a transceiver for transceiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations:

[0065] determining local training data related to a target application;

[0066] training a local machine learning model of the target application according to the local training data to obtain local model parameters of the trained local machine learning model, wherein the local machine learning model is obtained according to a global machine learning model trained by a first network function entity;

[0067] sending the local model parameters to the first network function entity, wherein the local model parameters are used to update the global machine learning model.

[0068] Optionally, the method further comprises:

[0069] obtaining global model parameters of the updated global machine learning model from the first network function entity, wherein the global model parameters are used to update the trained local machine learning model;

[0070] wherein the global model parameters are obtained by the first network function entity by updating the global machine learning model using at least one local model parameter sent by a UE.

[0071] Optionally, the sending the local model parameters to the first network function entity comprises:

[0072] sending model update information carrying the local model parameters to an application function (AF) entity through an application layer, so as to send the model update information to the first network function entity through the AF entity; or

[0073] sending model update information carrying the local model parameters to an application function (AF) entity through an application layer, so as to send the model update information to the first network function entity through the AF entity and a network exposure function (NEF) entity; or

[0074] The non-access layer sends model update information carrying the local model parameter to an access and mobility management function (AMF) entity, so as to send the model update information to the first network function entity through the AMF entity.

[0075] Optionally, the model update information further comprises any one or more of the following: a UE identifier, an application identifier, and a first network function entity identifier, for the AF entity or the AMF entity to determine the first network function entity and / or the global machine learning model.

[0076] Optionally, the obtaining of the global model parameter of the updated global machine learning model from the first network function entity comprises:

[0077] receiving a model update response sent by the AF entity or the AMF entity, wherein the model update response comprises information of the global model parameter of the updated global machine learning model.

[0078] Optionally, the local machine learning model is obtained according to the global machine learning model trained by the first network function entity, and the method comprises:

[0079] obtaining a first model file of the global machine learning model from the first network function entity;

[0080] creating the local machine learning model according to the first model file.

[0081] Optionally, the obtaining of the first model file of the global machine learning model from the first network function entity comprises:

[0082] sending model obtaining information carrying a UE identifier and / or model description information to the first network function entity;

[0083] obtaining information of the first model file of the global machine learning model from the first network function entity based on the model obtaining information.

[0084] Optionally, the sending of the model obtaining information carrying the UE identifier and / or the model description information to the first network function entity comprises:

[0085] sending model obtaining information carrying a UE identifier and / or model description information to the AF entity through an application layer, so as to send the model obtaining information to the first network function entity according to the UE identifier and / or the model description information through the AF entity; or

[0086] sending model obtaining information carrying a UE identifier and / or model description information to the AF entity through an application layer, so as to send the model obtaining information to the first network function entity according to the UE identifier and / or the model description information through the AF entity and a NEF entity; or

[0087] The non-access layer sends model obtaining information carrying UE identification and / or model description information to an AMF entity, so as to send the model obtaining information to the first network function entity according to the UE identification and / or model description information by the AMF entity.

[0088] Optionally, the obtaining the first model file of the global machine learning model from the first network function entity comprises:

[0089] receiving a model obtaining response sent by the AF entity or the AMF entity, wherein the model obtaining response comprises information of the first model file, and the first model file is obtained according to the information of the first model file.

[0090] Optionally, the model description information comprises at least one of the following:

[0091] application identification, application sub-feature identification, time information, location information, and other model feature information.

[0092] In a fifth aspect, the present application provides an AF entity, comprising a memory, a transceiver, and a processor:

[0093] The memory is configured to store a computer program; the transceiver is configured to transceive data under the control of the processor; and the processor is configured to read the computer program in the memory and perform the following operations:

[0094] receiving model update information carrying the local model parameters sent by the UE through an application layer, wherein the local machine learning model is a global machine learning model provided by a first network function entity to the UE in advance;

[0095] sending the model update information to the first network function entity, and the local model parameters are used to update the global machine learning model.

[0096] Optionally, the method further comprises:

[0097] obtaining information of global model parameters of the updated global machine learning model from the first network function entity, and the global model parameters are used to update the trained local machine learning model;

[0098] sending a model update response to the UE through an application layer, wherein the model update response comprises information of the global model parameters;

[0099] The global model parameters are obtained by the first network function entity by updating the global machine learning model using at least one local model parameter sent by the UE.

[0100] Optionally, the sending the model update information to the first network function entity comprises:

[0101] if the AF entity is a trusted AF entity, directly sending the model update information to the first network function entity; or

[0102] if the AF entity is an untrusted AF entity, sending the model update information to the first network function entity through a NEF entity.

[0103] Optionally, the obtaining the information of the global model parameters of the updated global machine learning model from the first network function entity comprises:

[0104] directly or through a NEF entity, obtaining the information of the global model parameters of the updated global machine learning model from the first network function entity.

[0105] Optionally, before the receiving the model update information carrying the local model parameters sent by the UE through an application layer, the method further comprises:

[0106] receiving model obtaining information carrying UE identification and / or model description information sent by the UE through an application layer;

[0107] sending the model obtaining information to the first network function entity according to the UE identification and / or model description information;

[0108] obtaining the information of a first model file of a global machine learning model from the first network function entity, and sending the information to the UE through an application layer.

[0109] Optionally, the sending the model obtaining information to the first network function entity according to the UE identification and / or model description information comprises:

[0110] if the AF entity is a trusted AF entity, selecting the first network function entity serving the UE and / or the first network function entity capable of providing a global machine learning model satisfying the model description information, and sending the model obtaining information to the first network function entity; or

[0111] if the AF entity is an untrusted AF entity, sending the model obtaining information to a NEF entity, selecting the first network function entity serving the UE and / or the first network function entity capable of providing a global machine learning model satisfying the model description information through the NEF entity, and sending the model obtaining information to the first network function entity.

[0112] Optionally, the information of the first model file of the global machine learning model obtained from the first network function entity is sent to the UE through an application layer, including:

[0113] The information of the first model file of the global machine learning model is obtained from the first network function entity directly or through a NEF entity;

[0114] The model obtaining response is sent to the UE through an application layer, wherein the information of the first model file is included in the model obtaining response, and the first model file is obtained according to the information of the first model file.

[0115] In a sixth aspect, the present application provides a first network function entity for providing a model, comprising a memory, a transceiver and a processor:

[0116] The memory is used for storing a computer program; the transceiver is used for transceiving data under the control of the processor; and the processor is used for reading the computer program in the memory and performing the following operations:

[0117] Obtaining local model parameters of a trained local machine learning model sent by a user equipment (UE), wherein the local machine learning model is obtained according to a global machine learning model trained by a first network function entity;

[0118] Updating the global machine learning model according to the local model parameters.

[0119] Optionally, after updating the global machine learning model according to the local model parameters, the method further comprises:

[0120] Sending global model parameters of the updated global machine learning model to the UE, wherein the global model parameters are used for updating the trained local machine learning model;

[0121] The global model parameters are obtained by updating the global machine learning model by using at least one local model parameter sent by the UE.

[0122] Optionally, before obtaining the local model parameters of the trained local machine learning model sent by the UE, the method further comprises:

[0123] Sending information of a first model file of the global machine learning model to the UE, so that the UE creates a local machine learning model according to the first model file.

[0124] In a seventh aspect, the present application provides a machine learning model processing device applied to a UE, comprising:

[0125] An obtaining unit is configured to determine local training data related to a target application;

[0126] A training unit is configured to train a local machine learning model of the target application according to the local training data, to obtain local model parameters of the trained local machine learning model, wherein the local machine learning model is obtained according to a global machine learning model trained by a first network function entity;

[0127] A sending unit is configured to send the local model parameters to the first network function entity, and the local model parameters are used to update the global machine learning model.

[0128] In an eighth aspect, the present application provides a machine learning model processing apparatus applied to an AF entity, and the apparatus comprises:

[0129] A receiving unit is configured to receive model update information carrying the local model parameters sent by a user equipment (UE) through an application layer, wherein the local machine learning model is a global machine learning model provided by a first network function entity to the UE in advance;

[0130] A sending unit is configured to send the model update information to the first network function entity, and the local model parameters are used to update the global machine learning model.

[0131] In a ninth aspect, the present application provides a machine learning model processing apparatus applied to a first network function entity for providing a model, and the apparatus comprises:

[0132] An obtaining unit is configured to obtain local model parameters of a trained local machine learning model sent by a user equipment (UE), wherein the local machine learning model is obtained according to a global machine learning model trained by a first network function entity;

[0133] A model updating unit is configured to update the global machine learning model according to the local model parameters.

[0134] In a tenth aspect, the present application provides a processor readable storage medium, which stores a computer program, and the computer program is used to make a processor execute the method in the first aspect or the second aspect or the third aspect.

[0135] In an eleventh aspect, the present application provides a computer program product, which comprises a computer program, and the computer program is used to make a processor execute the method in the first aspect or the second aspect or the third aspect.

[0136] The application provides a machine learning model processing method and device and a storage medium. A UE creates a local machine learning model of a target application according to a global machine learning model provided by a first network function entity in advance, determines local training data related to the target application, trains the local machine learning model according to the local training data, and sends local model parameters of the trained local machine learning model to the first network function entity to update the global machine learning model. The embodiments of the application can implement federated learning between the UE and the first network function entity for providing a model, improve the performance of sharing, transmitting and training the machine learning model between the UE and the network, and meet the requirements of rapidly developing communication services and applications.

[0137] It should be understood that the content described in the foregoing summary section is not intended to limit or define key or important features of the embodiments of the application, nor is it intended to limit the scope of the application. Other features of the application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0138] In order to more clearly illustrate the technical solutions in the application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0139] Figure 1 The application scenario schematic diagram of the machine learning model processing method provided by an embodiment of the application is shown in the figure.

[0140] Figure 2a The flowchart of the machine learning model processing method provided by an embodiment of the application is shown in the figure.

[0141] Figure 2b The flowchart of the machine learning model processing method provided by an embodiment of the application is shown in the figure.

[0142] Figure 3 The flowchart of the machine learning model processing method provided by another embodiment of the application is shown in the figure.

[0143] Figure 4 The signaling diagram of the machine learning model processing method provided by another embodiment of the application is shown in the figure.

[0144] Figure 5 The flowchart of the machine learning model processing method provided by another embodiment of the application is shown in the figure.

[0145] Figure 6 The signaling diagram of the machine learning model processing method provided by another embodiment of the application is shown in the figure.

[0146] Figure 7 A flowchart of a machine learning model processing method provided for another embodiment of the present application;

[0147] Figure 8 A signaling diagram of a machine learning model processing method provided for another embodiment of the present application;

[0148] Figure 9 A flowchart of a machine learning model processing method provided for another embodiment of the present application;

[0149] Figure 10 A signaling diagram of a machine learning model processing method provided for another embodiment of the present application;

[0150] Figure 11 A flowchart of a machine learning model processing method provided for another embodiment of the present application;

[0151] Figure 12 A signaling diagram of a machine learning model processing method provided for another embodiment of the present application;

[0152] Figure 13 A flowchart of a machine learning model processing method provided for another embodiment of the present application;

[0153] Figure 14 A signaling diagram of a machine learning model processing method provided for another embodiment of the present application;

[0154] Figure 15a A flowchart of a machine learning model processing method provided for another embodiment of the present application;

[0155] Figure 15b A flowchart of a machine learning model processing method provided for another embodiment of the present application;

[0156] Figure 16 A flowchart of a machine learning model processing method provided for another embodiment of the present application;

[0157] Figure 17 A structural diagram of a user equipment provided for an embodiment of the present application;

[0158] Figure 18 A structural diagram of an AF entity provided for an embodiment of the present application;

[0159] Figure 19 A structural diagram of a first network function entity provided for an embodiment of the present application;

[0160] Figure 20 A structural diagram of a machine learning model processing apparatus provided for an embodiment of the present application;

[0161] Figure 21A structural diagram of a machine learning model processing apparatus provided for another embodiment of the present application is shown in FIG. 1.

[0162] Figure 22 A structural diagram of a machine learning model processing apparatus provided for another embodiment of the present application is shown in FIG. 1. DETAILED DESCRIPTION

[0163] In this application, the term "and / or" describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. The character " / " generally represents an "or" relationship between the associated objects before and after it.

[0164] In the embodiments of this application, the term "multiple" refers to two or more, and other quantifiers are similar. In addition, the terms "first", "second", etc. are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features.

[0165] In order to realize intelligent data analysis and control of mobile communication networks, a network data analysis function (Network Data Analytics Function, NWDAF) is introduced in 5G networks. Based on artificial intelligence (Artificial Intelligence, AI) algorithms, NWDAF provides network data analysis (hereinafter referred to as "analysis" or "network analysis") services for other network function entities through interaction with other network function entities.

[0166] NWDAF can perform federated learning with other NWDAFs to obtain an optimized machine learning (Machine Learning, ML) model for AI inference. An existing process for federated learning between NWDAFs is as follows: one or more machine learning model consumer NWDAFs (ML model consumer) request a global model from a machine learning model provider NWDAF (ML model provider), then train the model using locally collected data to generate a local model, and send the locally updated model parameters to the machine learning model provider NWDAF. The machine learning model provider NWDAF updates the global model according to the local updated model parameters provided by the machine learning model consumer NWDAF, and then provides the updated global model to the machine learning model consumer NWDAF. This cycle is repeated multiple times to ultimately obtain a more optimal model for the network analysis.

[0167] The rapidly developing communication service and application demand requires user equipment (UE) to support AI function, and to realize the intelligentization of application and communication. The federated learning between UE and network will help to improve the training efficiency and performance of the machine learning model of UE and network, while protecting the privacy of UE local user data. However, the existing technology does not support the implementation of federated learning between UE and network (such as NWDAF), including UE subscription and update of machine learning model to network.

[0168] To solve the above technical problems, in the embodiments of the present application, the UE pre-creates a local machine learning model of a target application according to a first model file of a global machine learning model provided by a first network function entity, re-trains the local machine learning model according to local training data, sends the model related data of the trained local machine learning model to the first network function entity through an application layer or a non-access layer (Non-Access-Stratum, NAS), and updates the global machine learning model by the first network function entity; and the UE further acquires a second model file of the updated global machine learning model from the first network function entity through the application layer or the NAS, and updates the local machine learning model. The federated learning between the UE and the first network function entity for providing the model can be realized through the application layer or the non-access layer NAS, the performance of sharing, transmitting and training the machine learning model between the UE and the network is improved, and the rapidly developing communication service and application demand is met.

[0169] The embodiments of the present application are applied to an application scenario as shown in Figure 1 The application scenario includes a user equipment 101 and a first network function entity 102 for providing a model. Optionally, the first network function entity 102 can be a NWDAF entity. The user equipment 101 pre-creates a local machine learning model of a target application according to a first model file of a global machine learning model provided by the first network function entity 102, the first model file can include model parameters of the global machine learning model, and further, the user equipment 101 trains the local machine learning model according to local training data, sends local model parameters of the trained local machine learning model to the first network function entity 102 through an application layer or a non-access layer NAS, the first network function entity 102 updates the global machine learning model, and the user equipment 101 further acquires information of a second model file of the updated global machine learning model from the first network function entity through the application layer or the NAS, and updates the local machine learning model.

[0170] The user equipment (UE) involved in the embodiments of the present application can refer to a device providing voice and / or data connectivity for a user, a handheld device with wireless connection function, or other processing devices connected to a wireless modem, etc. In different systems, the name of the user equipment can also be different, for example, in the 5G system, the user equipment can be called UE. The wireless terminal device can communicate with the core network (CN) through the radio access network (RAN), and the wireless terminal device can be a mobile terminal device, such as a mobile phone (or called "cellular" phone) and a computer with a mobile terminal device, for example, it can be a portable, pocket-sized, handheld, computer-built-in or vehicle-mounted mobile device, which exchanges language and / or data with the radio access network. For example, personal communication service (PCS) phones, cordless phones, session initiated protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), etc. The wireless terminal device can also be called system, subscriber unit, subscriber station, mobile station, mobile, remote station, access point, remote terminal, access terminal, user terminal, user agent, user device, etc. in the embodiments of the present application, which are not limited.

[0171] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0172] The method and the device are based on the same application concept. Since the principles of the method and the device for solving problems are similar, the implementation of the device and the method can be referred to each other, and the repeated parts will not be described.

[0173] Embodiment one

[0174] Figure 2a A flowchart of the machine learning model processing method provided in this embodiment. As shown in the figure, this embodiment provides a machine learning model processing method, and the execution subject is a user equipment (UE). The specific steps of the method are as follows: Figure 2a

[0175] S201, determining local training data related to a target application;

[0176] S202, training a local machine learning model of the target application according to the local training data, to obtain local model parameters of the trained local machine learning model, wherein the local machine learning model is obtained according to a global machine learning model trained by a first network function entity;

[0177] S203, sending the local model parameters to the first network function entity, wherein the local model parameters are used to update the global machine learning model.

[0178] In this embodiment, the first network function entity for providing the model can provide a global machine learning model of the target application. In this embodiment, the specific machine learning model is not limited, for example, a machine learning model for image processing. Optionally, the first network function entity can be an NWDAF entity or other device capable of providing a machine learning model.

[0179] The UE can obtain a first model file of the global machine learning model from the first network function entity in advance, the first model file can include model parameters of the global machine learning model, and a local machine learning model of the target application can be created according to the first model file. Optionally, the UE can obtain the first model file from the first network function entity through an application layer or a non-access stratum (NAS). The specific process is not described here.

[0180] At any time after the UE creates the local machine learning model, for example, in the stage of applying the local machine learning model, the local machine learning model can be retrained based on local training data related to the target application in the UE, the model parameters of the local machine learning model are updated, and local model parameters are obtained. The specific retraining process can use any training method, which is not described here. The local training data can be any data related to the target application, for example, related data in the process of using the local machine learning model, or historical data of the target application, etc.

[0181] ​Further, after the local machine learning model is trained, in order to implement federated learning, the UE can send the local model parameters of the trained local machine learning model to the first network function entity, which can be sent to the first network function entity through the application layer or the non-access layer NAS, and the first network function entity can receive the local model parameters sent by any one or more UEs. In addition, the first network function entity can also receive the local model parameters sent by other network function entities (such as other NWDAFs), update the global machine learning model according to the received local model parameters, so as to isolate the data between UEs, protect user privacy and data security, and realize the learning and modeling of the global machine model. The process of updating the global machine learning model can adopt any model updating method in the federated learning scenario, which will not be described here.

[0182] In an optional embodiment, the UE can send the local model parameters of the trained local machine learning model to the first network function entity through the application layer, that is, the UE sends the local model parameters through the user plane. Specifically, the UE sends the model update information carrying the local model parameters to the AF (Application Function, application function) entity through the application layer, and then the AF entity sends the model update information carrying the local model parameters to the first network function entity directly or through the NEF (Network Exposure Function, network capability exposure function) entity, so that the first network function entity updates the global machine learning model according to the local model parameters.

[0183] In another optional embodiment, the UE can send the local model parameters of the trained local machine learning model to the first network function entity through the non-access layer NAS, that is, the UE sends the local model parameters through the control plane. Specifically, the UE sends the model update information including the local model parameters to the AMF (Access and Mobility Management Function, access and mobility management function) entity, and then the AMF entity can send the model update information including the local model parameters to the first network function entity, so that the first network function entity updates the global machine learning model according to the local model parameters.

[0184] The machine learning model processing method provided in the embodiment can realize federated learning between the UE and the first network function entity for providing the model, improve the performance of the UE and the network in sharing, transmitting and training the machine learning model, and meet the demand of rapidly developing communication services and applications.

[0185] Optionally, as shown in the above embodiment, after S203, the method can further include: Figure 2b

[0186] S204, obtaining global model parameters of the updated global machine learning model from the first network function entity, the global model parameters being used to update the trained local machine learning model.

[0187] The global model parameters are obtained by the first network function entity updating the global machine learning model by using the local model parameters sent by at least one UE.

[0188] In the embodiment, after the first network function entity updates the global machine learning model according to the received local model parameters, the first network function entity can transmit the global model parameters of the updated global machine learning model to the UE, so that the UE updates the local machine learning model according to the updated global model parameters, so that the updated local machine learning model has better prediction ability. Optionally, the first network function entity can transmit a second model file of the updated global machine learning model to the UE, or transmit address information of the second model file to the UE, wherein the second model file includes the global model parameters of the updated global machine learning model.

[0189] The UE can receive a model update response sent by the AF entity or the AMF entity, wherein the model update response includes information of the global model parameters of the updated global machine learning model.

[0190] ​Specifically, in an optional embodiment, the AF entity can obtain the global model parameter, or the second model file including the updated global model parameter, or address information of the second model file from the first network function entity, and send a model update response to the UE through an application layer, wherein the model update response includes the global model parameter, or the second model file, or the address information of the second model file, wherein the UE can obtain the second model file according to the address information of the second model file when obtaining the address information of the second model file, so as to obtain the updated global model parameter, and the model update response can further include subscription association information and UE identifier, etc., wherein the subscription association information can be the association information between the global model parameter and the UE.

[0191] In another optional embodiment, the AMF entity can obtain the global model parameter, or the second model file including the updated global model parameter, or address information of the second model file from the first network function entity, and send a model update response to the UE, wherein the model update response includes the global model parameter, or the second model file, or the address information of the second model file, wherein the UE can obtain the second model file according to the address information of the second model file when obtaining the address information of the second model file, so as to obtain the updated global model parameter, and the model update response can further include subscription association information, UE identifier, etc.

[0192] It should be noted that the first network function entity can provide a global machine learning model for a plurality of different UEs, therefore, the first network function entity can receive local model parameters of the trained local machine learning model sent by one or more UEs, and then update the global machine learning model according to the one or more local model parameters; and after the first network function entity updates the global machine learning model, the first network function entity can transmit a second model file of the updated global machine learning model to the one or more UEs, or to all UEs that subscribe to the global machine learning model.

[0193] Embodiment two

[0194] In this embodiment, the implementation of the local machine learning model obtained according to the global machine learning model trained by the first network function entity before S201 in Embodiment One is described in detail.

[0195] In this embodiment, the local machine learning model is obtained according to the global machine learning model trained by the first network function entity, and specifically includes:

[0196] Obtain a first model file of the global machine learning model from the first network function entity.

[0197] Create a local machine learning model according to the first model file.

[0198] In this embodiment, when the UE initially creates the local machine learning model, the UE can obtain a first model file of the global machine learning model from the first network function entity, wherein the first model file can include global model parameters of the global machine learning model, and then the UE can create the local machine learning model according to the global model parameters in the first model file, as a basis for the UE to train the local machine learning model according to local training data.

[0199] In this embodiment, the first model file of the global machine learning model obtained from the first network function entity includes:

[0200] The model acquisition information carrying the UE identifier and / or the model description information is sent to the first network function entity.

[0201] Based on the model acquisition information, the first model file information of the global machine learning model is obtained from the first network function entity.

[0202] In this embodiment, there can be at least one first network function entity providing different global machine models for the UE, so when the UE obtains the global machine model from the first network function entity, the UE can carry the UE identifier and / or the model description information, so as to select the first network function entity serving the UE and / or the first network function entity capable of providing the global machine learning model meeting the model description information according to the UE identifier and / or the model description information, and then obtain the global machine learning model required by the UE.

[0203] The model description information includes at least one of the following: application identifier, application sub-feature identifier, time information, location information, and other model feature information.

[0204] The application identifier (Application ID) is an application such as a camera application, a vehicle-to-everything (V2X) application, and a remote control.

[0205] The application sub-feature identifier (Feature ID) is an application sub-feature such as portrait photography in a camera application, navigation in a vehicle-to-everything application, and excavator control in a remote control.

[0206] The time information indicates a time period or a time point at which the machine learning model is used, such as a time period of camera shooting, vehicle-to-everything service, or remote control.

[0207] The location information indicates an area or a location at which the machine learning model is used, and can also include feature information of the area / location, such as city, countryside, mountain, and plain.

[0208] Other model feature information, such as resolution and filter information (black and white, retro, etc.) of the camera application, positioning / navigation accuracy of the vehicle networking service, and accuracy information of remote control.

[0209] On the basis of the above-mentioned embodiments, when sending the model obtaining information carrying the UE identifier and / or the model description information to the first network function entity, the following can be further included:

[0210] The model obtaining information carrying the UE identifier and / or the model description information is sent to the AF entity through the application layer, so that the AF entity sends the model obtaining information to the first network function entity according to the UE identifier and / or the model description information; or,

[0211] The model obtaining information carrying the UE identifier and / or the model description information is sent to the AF entity through the application layer, so that the AF entity sends the model obtaining information to the first network function entity according to the UE identifier and / or the model description information through the NEF entity; or,

[0212] The model obtaining information carrying the UE identifier and / or the model description information is sent to the AMF entity through the non-access layer, so that the AMF entity sends the model obtaining information to the first network function entity according to the UE identifier and / or the model description information.

[0213] In the present embodiment, the model obtaining information can be sent to the first network function entity in different ways as described above. The UE sends the model obtaining information to the AF entity through the application layer, which is based on the user plane to realize the transmission of the model obtaining information, while the non-access layer sends the model obtaining information to the AMF entity, which is based on the control plane to realize the transmission of the model obtaining information. Further, after the AF entity receives the model obtaining information, it can directly or through the NEF entity select the first network function entity according to the UE identifier and / or the model description information and then transmit the model obtaining information to the first network function entity. Similarly, after the AMF entity receives the model obtaining information, it can also select the first network function entity according to the UE identifier and / or the model description information and then transmit the model obtaining information to the first network function entity.

[0214] On the basis of the above-mentioned embodiments, the first model file of the global machine learning model is obtained from the first network function entity, which can include the following:

[0215] The model obtaining response sent by the AF entity or the AMF entity is received, wherein the model obtaining response includes information of the first model file, and the first model file is obtained according to the information of the first model file.

[0216] In this embodiment, similar to embodiment one, the AF entity can obtain the global model parameters, or the first model file including the global model parameters, or the address information of the first model file from the first network function entity, and send a model obtaining response to the UE through the application layer, wherein the model obtaining response includes the global model parameters, or the first model file, or the address information of the first model file, wherein the UE can obtain the first model file according to the address information of the first model file when obtaining the address information of the first model file, thereby obtaining the updated global model parameters, and in addition, the model obtaining response can further include subscription association information, UE identifier, etc., wherein the subscription association information can be the association information between the global model parameters and the UE.

[0217] In another optional embodiment, the AMF entity can obtain the global model parameters, or the first model file including the global model parameters, or the address information of the first model file from the first network function entity, and send a model obtaining response to the UE, wherein the model obtaining response includes the global model parameters, or the first model file, or the address information of the first model file, wherein the UE can obtain the first model file according to the address information of the first model file when obtaining the address information of the first model file, thereby obtaining the updated global model parameters, and in addition, the model obtaining response can further include subscription association information, UE identifier, etc.

[0218] Embodiment three

[0219] In this embodiment, the implementation of the UE obtaining the local machine learning model according to the global machine learning model trained by the first network function entity in embodiment two is introduced in detail.

[0220] In this embodiment, the UE can obtain the information of the first model file of the global machine learning model from the first network function entity through the application layer, wherein the information of the first model file can include the global model parameters, or the first model file including the global model parameters, or the address information of the first model file, and then create the local machine learning model according to the information of the first model file, that is, the UE requests the subscription of the global machine learning model from the first network function entity through the application layer to the AF entity, and the AF entity obtains the first model file of the global machine learning model from the first network function entity on behalf of the UE and provides it to the UE through the application layer, thereby realizing the machine learning model subscription in the user plane.

[0221] The AF entity in this embodiment is a trusted AF entity, as shown in Figure 3 The above process is as follows:

[0222] S301, sending, by an application layer, a first model obtaining request to an AF entity, so that the AF entity determines a first network function entity according to the first model obtaining request, and obtains information of a first model file of a global machine learning model from the first network function entity, wherein the first model obtaining request comprises a UE identifier and / or model description information;

[0223] S302, receiving a model obtaining response sent by the AF entity through the application layer;

[0224] The model obtaining response can comprise information of the first model file, and the information of the first model file can comprise global model parameters, or a first model file comprising the global model parameters, or address information of the first model file. The model obtaining response can further comprise subscription association information, a UE identifier, etc. The information of the first model file is obtained by the AF entity from the first network function entity.

[0225] S303, creating a local machine learning model according to the information of the first model file.

[0226] In this embodiment, the UE requests the first network function entity to obtain the global machine learning model through the AF entity by means of the application layer. After the AF entity obtains the global machine learning model, the AF entity can send the global machine learning model to the UE through the application layer.

[0227] Optionally, before S301, a PDU (Protocol Data Unit) session between the UE and the AF entity can be established first, which is used to connect the UE and the AF entity in the application layer to transmit the first model obtaining request and the model obtaining response.

[0228] The PDU session comprises a PDU session dedicated to transmitting machine learning model data, or a PDU session dedicated to carrying specific application data. The PDU session comprises a QoS flow dedicated to transmitting machine learning model data, and / or a QoS flow dedicated to transmitting machine learning model related signaling, wherein the QoS flow is established through a PDU session establishment or modification process.

[0229] On the basis of the above embodiment, the first model obtaining request comprises a UE identifier and / or model description information. The AF entity can select an alternative first network function entity serving the UE and / or an alternative first network function entity capable of providing a global machine learning model satisfying the model description information as the first network function entity according to the UE identifier and / or the model description information. The model description information comprises at least one of the following: an application identifier, an application sub-feature identifier, time information, location information, and other model feature information.

[0230] wherein, Application ID, the application is, for example, a camera application, a vehicle-to-everything (V2X: Vehicle to Everything) application, a remote control;

[0231] Application sub-feature ID, application sub-feature, for example, portrait photography in the camera application, navigation in the vehicle-to-everything application, excavator control in the remote control;

[0232] Time information, indicating the time period or time point used by the machine learning model, for example, the time period of camera shooting, vehicle-to-everything service or remote control;

[0233] Location information, indicating the area or location used by the machine learning model, and can also contain feature information of the area / location, such as city, countryside, mountain, plain, etc.

[0234] Other model feature information, such as resolution and filter information (black and white, retro, etc.) of the camera application, positioning / navigation accuracy of the vehicle-to-everything service, and accuracy information of the remote control.

[0235] As an example, the above process of the embodiment is described in detail below in combination with the signaling diagram as shown in Figure 4

[0236] S311, establishing a PDU session between the UE and the AF entity to connect the UE and the AF entity at the application layer;

[0237] Wherein, the PDU session includes: a PDU session dedicated to transmitting machine learning model data, or a PDU session dedicated to carrying specific application data.

[0238] S312, the UE sends a first model acquisition request for the global machine learning model to the AF entity through the application layer;

[0239] Wherein, the first model acquisition request can include UE identifier and / or model description information;

[0240] Optionally, the first model acquisition request can also include the NWDAF entity identifier, indicating the NWDAF entity serving the UE, so that the AF entity can directly determine the NWDAF entity without performing the process of selecting the NWDAF entity described below; the NWDAF entity identifier can be preconfigured in the UE or determined through the historical model subscription and update process.

[0241] S313, the AF entity selects the NWDAF entity according to the first model acquisition request;

[0242] ​The AF entity can select an alternative NWDAF entity serving the UE and / or an alternative NWDAF entity capable of providing a global machine learning model satisfying the model description information, and determine the alternative NWDAF entity as the final NWDAF entity.

[0243] In S314, the AF entity sends a second model obtaining request to the NWDAF entity.

[0244] The second model obtaining request includes the UE identifier and / or the model description information. The AF entity can extract the required information from the first model obtaining request to generate the second model obtaining request sent to the NWDAF.

[0245] In S315, the NWDAF entity generates subscription correlation information, determines a global machine learning model according to the second obtaining request, and sends a model obtaining notification to the AF entity.

[0246] The model subscription is an optional way of model obtaining. After the UE subscribes to a global machine learning model, the first network function entity can push the updated global model parameters of the global machine learning model to the UE after the global machine learning model is updated. Specifically, the model subscription notification includes model information of the global machine learning model and subscription correlation information. The model information includes information of a first model file. The information of the first model file can be the following information: global model parameters of the global machine learning model, or a first model file containing the global model parameters of the global machine learning model, or address information of the first model file, or information of a network function entity (such as a database function entity) storing the first model file, etc. The subscription correlation information can be a subscription correlation identifier (Subscription Correlation ID). The NWDAF entity assigns the subscription correlation identifier. Specifically, the NWDAF entity can assign the same subscription correlation identifier to different UEs subscribing to the same model, or assign different subscription correlation identifiers to different UEs subscribing to the same model. Of course, the model subscription mode can not be used in the example, i.e., the above-mentioned subscription correlation information can not be involved.

[0247] The NWDAF entity can provide a global machine learning model satisfying the model description information. For example, the model description information is black-and-white portrait photography in a rural environment from 6 pm to 7 pm. The model provided by the NWDAF entity can achieve the best / better photography effect in this scenario. Or, the model description information is traffic navigation information in region A from 6 pm to 7 pm. The model provided by the NWDAF entity can achieve the best / better navigation effect in this time and region.

[0248] Optionally, in S316, the AF entity obtains the first model file according to the information of the first model file.

[0249] S317, the AF entity sends a model obtaining response to the UE through the application layer;

[0250] The model obtaining response includes information of the first model file, and can also include subscription association information and a UE identifier.

[0251] S318, the UE obtains the first model file according to the information of the first model file, and creates a local machine learning model according to the first model file. That is, the UE can create a local machine learning model based on the global model parameters in the first model file.

[0252] Through the above process, the UE can obtain a machine learning model from a first network function entity for providing a model through the application layer, thereby improving the performance of sharing, transmitting, and training a machine learning model between the UE and the network, and meeting the requirements of rapidly developing communication services and applications.

[0253] Embodiment Four

[0254] In this embodiment, the implementation of the UE obtaining a local machine learning model according to a global machine learning model trained by a first network function entity in Embodiment Two is described in detail.

[0255] In this embodiment, the UE can obtain information of a first model file of a global machine learning model from the first network function entity through the application layer. The information of the first model file can include global model parameters, or a first model file including global model parameters, or address information of the first model file. Then, the UE creates a local machine learning model according to the information of the first model file. That is, the UE requests the first network function entity for obtaining a global machine learning model through the application layer to the AF entity, the AF entity obtains a first model file of a global machine learning model from the first network function entity on behalf of the UE, and provides the first model file to the UE through the application layer, thereby realizing machine learning model obtaining in the user plane.

[0256] Different from the above embodiments, the AF entity in this embodiment is an untrusted AF entity. The AF entity can use the NEF to realize the above process, as shown in FIG. 4. Figure 5 The specific process is as follows.

[0257] S401, a first model obtaining request is sent to the AF entity through the application layer, so that the AF entity determines a first network function entity through the NEF entity according to the first model obtaining request, and then obtains information of a first model file of a global machine learning model from the first network function entity. The first model obtaining request includes a UE identifier and / or model description information.

[0258] S402, a model obtaining response sent by the AF entity through the application layer is received,

[0259] The model obtaining response includes information of the first model file, and the information of the first model file can include global model parameters or a first model file including the global model parameters or address information of the first model file. The model obtaining response can further include subscription association information, a UE identifier, and the like. The information of the first model file is obtained by the AF entity from the first network function entity;

[0260] S403, creating a local machine learning model according to the information of the first model file.

[0261] In this embodiment, since the AF entity is not trusted, the communication between the AF and the first network function entity is implemented through the NEF entity, that is, the UE sends a first model obtaining request to the AF entity through an application layer, the AF entity sends a third model obtaining request to the NEF entity, the NEF entity selects a first network function entity and sends a fourth model obtaining request to the first network function entity, and the first network function entity can send the information of the first model file to the AF through the NEF entity. After the AF entity obtains the information of the first model file, the AF entity can send the information of the first model file to the UE through the application layer.

[0262] Optionally, before S401, a PDU session between the UE and the AF entity can be established to connect the UE and the AF entity through an application layer to transmit the first model obtaining request and the model obtaining response, which can be referred to in Embodiment Two.

[0263] On the basis of the above embodiments, the first model obtaining request includes a UE identifier and / or model description information. The NEF entity can select an alternative first network function entity serving the UE and / or an alternative first network function entity capable of providing a global machine learning model satisfying the model description information according to the UE identifier and / or the model description information, and determine the first network function entity, which can be referred to in Embodiment Two.

[0264] As an example, the above process of this embodiment is described in detail in combination with a signaling diagram as shown in FIG. 4, in which the first network function entity takes the NWDAF entity as an example. Figure 6

[0265] S411, establishing a PDU session between the UE and the AF entity to connect the UE and the AF entity through an application layer;

[0266] S412, the UE sends a first model obtaining request for a global machine learning model to the AF entity through an application layer;

[0267] The first model obtaining request can include a UE identifier and / or model description information.

[0268] ​Optionally, the first model acquisition request can also include an NWDAF entity identifier, indicating the NWDAF entity serving the UE, so that the NEF entity can directly determine the NWDAF entity without performing the following process of selecting the NWDAF entity.

[0269] S413, the AF entity sends a third model acquisition request to the NEF entity according to the first model acquisition request;

[0270] The third model acquisition request can include a UE identifier and / or model description information.

[0271] S414, the NEF entity selects an NWDAF entity according to the third model acquisition request; the AF entity can generate a third model acquisition request that can be sent to the NEF after extracting the required information from the first model acquisition request.

[0272] The NEF entity can select an alternative NWDAF entity serving the UE and / or an alternative NWDAF entity capable of providing a global machine learning model satisfying the model description information, and determine the final NWDAF entity.

[0273] S415, the NEF entity sends a fourth model acquisition request to the NWDAF entity;

[0274] The fourth model acquisition request includes a UE identifier and / or model description information.

[0275] S416, the NWDAF entity generates subscription association information, determines a global machine learning model according to the fourth acquisition request, and sends a model acquisition notification to the NEF entity;

[0276] The model acquisition notification includes model information of the global machine learning model and subscription association information. The model information includes information of a first model file. The information of the first model file can be the following information: a first model file containing parameters of the global machine learning model, an address of the first model file, or information of a network function entity (such as a database function entity) storing the first model file, etc. The subscription association information can be a subscription association identifier, which is assigned by the NWDAF entity. Specifically, the NWDAF entity can assign the same subscription association identifier to different UEs subscribing to the same model, or assign different subscription association identifiers to different UEs subscribing to the same model. Of course, the model subscription method can not be used in this example, that is, the above-mentioned subscription association information can not be involved.

[0277] S417, the NEF entity sends a model acquisition notification to the AF entity;

[0278] The model obtaining notification includes model information of the global machine learning model and subscription association information. The model information includes information of a first model file. The information of the first model file can be the following information: a first model file containing parameters of the global machine learning model, an address of the first model file, or information of a network function entity (for example, a database function entity) storing the first model file, etc.

[0279] Optionally, in S418, the AF entity obtains the first model file according to the information of the first model file.

[0280] In S419, the AF entity sends a model obtaining response to the UE through an application layer.

[0281] The model obtaining response includes the information of the first model file, and can also include subscription association information and a UE identifier.

[0282] In S4110, the UE obtains the first model file according to the information of the first model file, and creates a local machine learning model according to the first model file. That is, the UE can create a local machine learning model based on the global model parameters in the first model file.

[0283] Through the above process, the UE can obtain a machine learning model from a first network function entity for providing a model through an application layer, thereby improving the performance of sharing, transmitting, and training a machine learning model between the UE and the network, and meeting the requirements of rapidly developing communication services and applications.

[0284] Embodiment Five

[0285] In this embodiment, the implementation of the UE obtaining a local machine learning model according to a global machine learning model trained by the first network function entity in Embodiment Two is described in detail.

[0286] In this embodiment, the UE can obtain information of a first model file of the global machine learning model from the first network function entity through a non-access layer NAS. The information of the first model file can include global model parameters, or a first model file containing the global model parameters, or address information of the first model file. Then, the UE creates a local machine learning model according to the information of the first model file. Figure 7 As shown in the following figure, the specific process is as follows:

[0287] In S501, a third NAS message for obtaining a model is sent to an AMF entity, so that the AMF entity obtains information of a first model file of a global machine learning model from a first network function entity according to the third NAS message.

[0288] The information of the first model file can be the following information: global model parameters, or a first model file containing global machine learning model parameters, or address information of the first model file, or information of a network function entity (for example, a database function entity) storing the first model file, etc.

[0289] S502, receiving a fourth NAS message for acquiring a model sent by the AMF entity, wherein the fourth NAS message includes information of the first model file and acquisition association information;

[0290] S503, acquiring the first model file according to the information of the first model file;

[0291] S504, creating a local machine learning model according to the first model file.

[0292] In this embodiment, the UE requests the first network function entity to acquire the global machine learning model through the AMF entity by means of the non-access layer NAS of the control plane. After the AMF entity acquires the information of the first model file of the global machine learning model, the AMF entity can send the information to the UE through the non-access layer NAS. The UE can acquire the global machine learning model according to the information of the first model file.

[0293] On the basis of the above-mentioned embodiments, the third NAS message includes a UE identifier and / or model description information; wherein the AMF entity can select an alternative first network function entity serving the UE and / or an alternative first network function entity capable of providing a global machine learning model satisfying the model description information according to the UE identifier and / or the model description information, and determine the first network function entity, which can be referred to Embodiment 2.

[0294] Optionally, the NAS message sent by the UE to the AMF entity can include but is not limited to an uplink non-access layer transmission (UL NAS Transport) message, a registration request (Registration Request) message, a service request (Service Request) message, etc.

[0295] As an example, the above-mentioned process of this embodiment will be introduced in detail in combination with a signaling diagram as shown in FIG. 5. Figure 8

[0296] S511, sending a third NAS message for acquiring a model to the AMF entity;

[0297] The third NAS message includes a UE identifier and / or model description information;

[0298] ​Optionally, the NWDAF entity identifier can also be included in the third NAS message, indicating the NWDAF entity serving the UE, so that the AMF entity can directly determine the NWDAF entity without performing the following process of selecting the NWDAF entity.

[0299] S512, the AMF entity selects the NWDAF entity according to the third NAS message;

[0300] Among them, the AMF entity can select the alternative NWDAF entity serving the UE and / or the alternative NWDAF entity capable of providing a global machine learning model meeting the model description information, and determine it as the final NWDAF entity.

[0301] S513, the AMF entity sends a fifth model acquisition request to the NWDAF entity;

[0302] Among them, the fifth model acquisition request includes the UE identifier and / or the model description information; the AMF entity can extract the required information from the third NAS message request to generate the fifth model acquisition request that can be sent to the NWDAF, of course, the AMF entity can transmit the third NAS message request to the NWDAF as the fifth model acquisition request.

[0303] S514, the NWDAF entity generates subscription association information, determines the global machine learning model according to the fifth acquisition request, and sends a model acquisition notification to the AMF entity;

[0304] Among them, the model acquisition notification includes the model information of the global machine learning model and the subscription association information, and the model information includes the information of the first model file; the information of the first model file can be the following information: the global model parameter of the global machine learning model, or the first model file containing the global machine learning model parameter, or the address information of the first model file, or the information of the network function entity (such as the database function entity) storing the first model file, etc.

[0305] The subscription association information can be a subscription association identifier, and the NWDAF entity allocates the subscription association identifier; specifically, the NWDAF entity can allocate the same subscription association identifier to different UEs subscribing to the same model, or allocate different subscription association identifiers to different UEs subscribing to the same model. Of course, the model subscription method can also not be used in this example, that is, the above-mentioned subscription association information can not be involved.

[0306] S515, the AMF entity sends a fourth NAS message for acquiring the model to the UE;

[0307] Among them, the fourth NAS message includes the information of the first model file and the subscription association information;

[0308] S516, the UE acquires the first model file according to information of the first model file;

[0309] S517, the UE creates a local machine learning model according to the first model file. That is, the UE can create a local machine learning model based on the global model parameters in the first model file.

[0310] The above process can be used to implement the UE to acquire the machine learning model from the first network function entity for providing the model through the non-access stratum (NAS), improve the performance of sharing, transmitting and training the machine learning model between the UE and the network, and meet the requirements of the rapidly developing communication services and applications.

[0311] Embodiment six

[0312] In this embodiment, the possible implementation of S201-S204 in Embodiment One is described in detail.

[0313] In this embodiment, the UE can transmit the local model parameters to the first network function entity through the AF entity of the application layer after retraining the local machine learning model, and transmit the information of the second model file of the updated global machine learning model to the UE through the AF entity of the application layer after the first network function entity updates the global machine learning model, so as to realize the update of the machine learning model in the user plane. The information of the second model file can include the updated global model parameters, or a second model file including the updated global model parameters, or address information of the second model file.

[0314] The AF entity in this embodiment is a trusted AF entity, as shown in Figure 9 After the UE creates the local machine learning model and acquires the local training data, the above process is as follows:

[0315] S601, retrain the local machine learning model of the target application according to the local training data;

[0316] S602, send a first model update request to an application function (AF) entity through an application layer, so that the AF entity requests the first network function entity to update the global machine learning model according to the first model update request, and the first model update request includes the local model parameters;

[0317] S603, receive a model update response sent by the AF entity through the application layer;

[0318] The model update response includes information of a second model file of the updated global machine learning model, the information of the second model file can include updated global model parameters, or a second model file including the updated global model parameters, or address information of the second model file, and the model update response can further include subscription association information, a UE identifier, etc. The information of the second model file is obtained by the AF entity from the first network function entity;

[0319] S604, updating the local machine learning model according to the information of the second model file.

[0320] Specifically, the updated global model parameters included in the second model file are obtained according to the information of the second model file, and the local machine learning model is updated according to the updated global model parameters.

[0321] In this embodiment, the UE sends the local model parameters to the first network function entity through the application layer via the AF entity, so that the first network function entity updates the global machine learning model, and the AF entity can send the updated global machine learning model to the UE through the application layer.

[0322] Optionally, before S601, a PDU session between the UE and the AF entity can be established first, which is used to connect the UE and the AF entity through the application layer to transmit the first model update request and the model update response. Of course, the PDU session established in the model obtaining stage can also be used.

[0323] Optionally, the first model update request further includes at least one of the following: subscription association information, a UE identifier, an application identifier, and a first network function entity identifier, which are used by the AF entity to determine the first network function entity and / or the global machine learning model. In this embodiment, the model subscription context of the UE can be stored when the model is subscribed, and then the first network function entity providing the global machine learning model for the UE is determined based on the model subscription context through the subscription association information, the UE identifier, the application identifier, and the first network function entity identifier.

[0324] As an example, the above process of this embodiment will be described in detail below in combination with a signaling diagram as shown in FIG. 6, in which the first network function entity takes the NWDAF entity as an example. Figure 10

[0325] S611, the UE re-trains the local machine learning model of the target application according to the local training data;

[0326] S612, the UE sends a first model update request to the AF entity through the application layer;

[0327] ​The first model update request comprises the local model parameter and at least one of the following: subscription association information, UE identifier, application identifier, and NWDAF identifier.

[0328] S613, the AF entity determines the NWDAF according to the first model update request, and sends a second model update request to the NWDAF.

[0329] The second model update request can comprise the local model parameter. The AF entity can extract the required information from the first model update request and generate the second model update request that can be sent to the NWDAF.

[0330] S614, the NWDAF updates the global machine learning model according to the local model parameter.

[0331] S615, the NWDAF sends a model update completion notification to the AF entity.

[0332] The model update completion notification comprises model information of the updated global machine learning model and subscription association information. The model information comprises information of a second model file. The information of the second model file can be the following information: updated global model parameter, or a second model file containing the updated global model parameter, or address information of the second model file, or information of a network function entity (for example, a database function entity) storing the second model file, etc.

[0333] Optionally, S616, the AF entity acquires the second model file according to the information of the second model file.

[0334] S617, the AF entity sends a model update response to the UE through the application layer.

[0335] The model update response comprises the information of the second model file, and can further comprise subscription association information and UE identifier.

[0336] S618, the UE acquires the second model file according to the information of the second model file, and updates the local machine learning model according to the second model file. That is, the UE can update the local machine learning model based on the updated global model parameter in the second model file.

[0337] The above process can realize federated learning of machine learning models between the UE and the NWDAF for providing the models through the application layer, improve the performance of sharing, transmitting and training machine learning models between the UE and the network, and meet the rapidly developing communication services and application requirements.

[0338] Embodiment seven

[0339] In this embodiment, possible implementation manners of S201-S204 in Embodiment One are described in detail.

[0340] In this embodiment, after retraining the local machine learning model, the UE can transmit the local model parameters to the first network function entity through the AF entity of the application layer, and after the first network function entity updates the global machine learning model, the UE transmits the information of the second model file of the updated global machine learning model to the UE through the AF entity of the application layer, so as to realize the update of the machine learning model in the user plane. The information of the second model file can include the updated global model parameters, or the second model file including the updated global model parameters, or the address information of the second model file. Different from the above embodiment, the AF entity in this embodiment is an untrusted AF entity, and the AF entity can use the NEF to realize the above process, such as Figure 11 After the UE creates the local machine learning model and obtains the local training data, the above process is as follows:

[0341] S701, retrain the local machine learning model of the target application according to the local training data;

[0342] S702, send a first model update request to an application function AF entity through an application layer, so that the AF entity requests the first network function entity to update the global machine learning model through the NEF entity according to the first model update request, and the first model update request includes the local model parameters;

[0343] S703, receive a model update response sent by the AF entity through the application layer,

[0344] The model update response includes the information of the second model file of the updated global machine learning model, and the information of the second model file can include the updated global model parameters, or the second model file including the updated global model parameters, or the address information of the second model file. The model update response can also include subscription association information, UE identifier, etc. The information of the second model file is obtained by the AF entity from the first network function entity;

[0345] S704, update the local machine learning model according to the information of the second model file.

[0346] Specifically, the updated global model parameters included in the second model file are obtained according to the information of the second model file, and the local machine learning model is updated according to the updated global model parameters included in the second model file.

[0347] In this embodiment, since the AF entity is not trusted, the communication between the AF and the first network function entity is implemented through the NEF entity, that is, the UE sends a first model update request to the AF entity through the application layer, the AF entity sends a third model update request to the NEF entity, the NEF entity determines the first network function entity and sends a fourth model update request to the first network function entity; and the first network function entity can send the information of the second model file to the AF through the NEF entity, and the AF entity can send the information of the second model file to the UE through the application layer after obtaining the information of the second model file.

[0348] Optionally, before S601, a PDU session between the UE and the AF entity can be established first, which is used to connect the UE and the AF entity through the application layer to transmit the first model update request and the model update response. Of course, the PDU session established in the model obtaining stage can also be used.

[0349] Optionally, the first model update request further includes at least one of the following: subscription association information, UE identifier, application identifier, and first network function entity identifier, which are used for the NEF entity to determine the first network function entity and / or the global machine learning model. In this embodiment, the model subscription context of the UE can be stored when the model is subscribed, and then the first network function entity providing the global machine learning model for the UE is determined based on the model subscription context through the subscription association information, UE identifier, application identifier, and first network function entity identifier.

[0350] As an example, the above process of this embodiment will be described in detail in combination with the signaling diagram as shown in FIG. 7. Figure 12

[0351] S711, the UE re-trains the local machine learning model of the target application according to the local training data;

[0352] S712, the UE sends a first model update request to the AF entity through the application layer;

[0353] The first model update request includes the local model parameter and at least one of the following: subscription association information, UE identifier, application identifier, and NWDAF identifier.

[0354] S713, the AF entity sends a third model update request to the NEF entity;

[0355] The third model update request can include the local model parameter and at least one of the following: subscription association information, UE identifier, application identifier, and NWDAF identifier. The AF entity can extract the required information from the first model update request to generate the third model update request that can be sent to the NEF.​

[0356] S714, the NEF entity determines the NWDAF entity according to the third model update request, and sends a fourth model update request to the NWDAF entity;

[0357] The fourth model update request includes subscription association information and the local model parameter; the NEF entity can extract the required information from the third model update request to generate the fourth model update request that can be sent to the NWDAF.

[0358] S715, the NWDAF entity updates the global machine learning model according to the local model parameter;

[0359] S716, the NWDAF entity sends a model update completion notification to the NEF entity;

[0360] The model update completion notification includes model information of the updated global machine learning model and subscription association information, and the model information includes information of the second model file; the information of the second model file can be the following information: the updated global model parameter, or the second model file containing the updated global model parameter, or address information of the second model file, or information of a network function entity (such as a database function entity) storing the second model file, etc.

[0361] S717, the NEF entity sends a model update completion notification to the AF entity;

[0362] The model update completion notification includes model information of the updated global machine learning model of the NWDAF entity, and can also include subscription association information, and the model information includes information of the second model file;

[0363] Optionally, S718, the AF entity acquires the second model file according to the information of the second model file;

[0364] S719, the AF entity sends a model update response to the UE through the application layer;

[0365] The model update response includes information of the second model file, and can also include subscription association information and UE identifier;

[0366] S7110, the UE acquires the second model file according to the information of the second model file, and updates the local machine learning model according to the second model file. That is, the UE can update the local machine learning model based on the updated global model parameter in the second model file.

[0367] The embodiment can improve the performance of sharing, transmitting and training the machine learning model between the UE and the network by implementing the federated learning of the machine learning model between the UE and the first network function entity for providing the model through the application layer, and meets the demand of the rapidly developing communication service and application.

[0368] Embodiment eight

[0369] The possible implementation of S201-S204 in embodiment one is described in detail in this embodiment.

[0370] In this embodiment, the UE can transmit the local model parameters to the first network function entity through the AMF entity of the non-access layer NAS after retraining the local machine learning model, and transmit the information of the second model file of the updated global machine learning model to the UE through the AMF entity of the non-access layer NAS after the first network function entity updates the global machine learning model, so as to realize the update of the machine learning model in the user plane, such as Figure 13 After the UE creates the local machine learning model and obtains the local training data, the above process is as follows:

[0371] S801, retrain the local machine learning model of the target application according to the local training data;

[0372] S802, send a first NAS message of requesting to update the model to an access and mobility management function AMF entity, so that the AMF entity requests the first network function entity to update the global machine learning model according to the first NAS message, and the first NAS message includes the local model parameters;

[0373] S803, receive a second NAS message of responding to the model update sent by the AMF entity;

[0374] The second NAS message includes the information of the second model file, and the information of the second model file can include the updated global model parameters, or the second model file including the updated global model parameters, or the address information of the second model file, and the model update response can also include the subscription association information and the like;

[0375] S804, obtain the second model file according to the information of the second model file;

[0376] S805, update the local machine learning model according to the second model file.

[0377] Specifically, the updated global model parameters included in the second model file are obtained according to the information of the second model file, and the local machine learning model is updated according to the updated global model parameters included in the second model file.

[0378] In this embodiment, the UE sends local model parameters to the first network function entity via the Non-Access Stratum (NAS) entity of the control plane using the AMF entity, so that the first network function entity updates the global machine learning model. After updating the global machine learning model, the first network function entity can send the information of the second model file to the UE via the NAS entity using the AMF entity. The second model file contains the updated global model parameters.

[0379] Optionally, the first NAS message may further include at least one of the following: subscription association information, UE identifier, application identifier, and first network function entity identifier, for use by the AMF entity to determine the first network function entity and / or the global machine learning model. In this embodiment, the AMF may store the UE's model subscription context during model subscription, and then, based on the model subscription context, determine the first network function entity that provides the global machine learning model to the UE through the subscription association information, UE identifier, application identifier, and first network function entity identifier.

[0380] Optionally, the NAS messages sent by the UE to the AMF entity may include, but are not limited to, uplink non-access stratum transport (ULNAS Transport) messages, registration request messages, service request messages, etc.

[0381] As an example, the following is combined with... Figure 14 The signaling diagram shown below provides a detailed description of the above process in this embodiment, with the NWDAF entity serving as an example of the first network functional entity.

[0382] S811, UE retrains the local machine learning model of the target application based on local training data;

[0383] S812, The UE sends the first NAS message requesting an update to the model to the AMF entity;

[0384] The first model update request includes the local model parameters and at least one of the following: subscription association information, UE identifier, application identifier, and NWDAF identifier;

[0385] S813, the AMF entity determines the NWDAF entity based on the first NAS message and sends the fifth model update request to the NWDAF entity;

[0386] The fifth model update request includes local model parameters, and may also include subscription-related information, etc.

[0387] S814, NWDAF updates the global machine learning model based on the local model parameters;

[0388] S815 and NWDAF send a notification to the AMF entity that the model update is complete;

[0389] The notification that the model update is complete includes the updated global machine learning model information and subscription association information. The model information includes the information of the second model file. The information of the second model file may be the following: the updated global model parameters, or the address information of the second model file, or the information of the network functional entity (e.g., the database functional entity) storing the second model file.

[0390] S816, the AMF entity sends a second NAS message to the UE in response to the model update;

[0391] The second NAS message includes information about the second model file and may also include subscription association information;

[0392] S817. The UE obtains the second model file based on the information in the second model file;

[0393] S818, the UE updates its local machine learning model based on the second model file. That is, the UE can update its local machine learning model based on the updated global model parameters in the second model file.

[0394] This embodiment, through the above process, enables federated learning of machine learning models between the UE and the NWDAF used to provide models through the non-access stratum (NAS), improving the performance of sharing, transmitting, and training machine learning models between the UE and the network, and meeting the needs of rapidly developing communication services and applications.

[0395] Example 9

[0396] Figure 15a This is a flowchart of the machine learning model processing method provided in this embodiment. Figure 15a As shown, this embodiment provides a machine learning model processing method, with the AF entity as the execution subject. The specific steps of this machine learning model processing method are as follows:

[0397] S901. Receive model update information carrying the local model parameters sent by the UE through the application layer, wherein the local machine learning model is a global machine learning model provided to the UE in advance by the first network function entity.

[0398] S902. The model update information is sent to the first network function entity, and the local model parameters are used to update the global machine learning model.

[0399] Based on the above embodiments, such as Figure 15b As shown, S902 may also include:

[0400] S903. Obtain information on the global model parameters of the updated global machine learning model from the first network functional entity. The global model parameters are used to update the trained local machine learning model.

[0401] S904. Send a model update response to the UE through the application layer, wherein the model update response includes information about the global model parameters;

[0402] The global model parameters are obtained by the first network function entity updating the global machine learning model using local model parameters sent by at least one UE.

[0403] The federated learning method for mobile networks provided in this embodiment is the AF entity-side method in the above embodiments. Its principle and technical effects can be found in the above embodiments, and will not be repeated here.

[0404] Based on the above embodiments, sending the model update information to the first network functional entity may specifically include:

[0405] If the AF entity is a trusted AF entity, then the model update information is directly sent to the first network function entity; or

[0406] If the AF entity is an untrusted AF entity, the model update information is sent to the first network function entity through the NEF entity.

[0407] Furthermore, obtaining the updated global model parameters of the global machine learning model from the first network functional entity includes:

[0408] The updated global model parameters of the global machine learning model can be obtained directly or through the NEF entity from the first network functional entity.

[0409] In this embodiment, the model update information may be a first model update request sent by the UE through the application layer. The first model update request includes the local model parameters and at least one of the following: subscription association information, UE identifier, application identifier, and first network function entity identifier.

[0410] In one optional embodiment, the AF entity is a trusted AF entity; the AF entity can determine the first network function entity based on the first model update request; and send a second model update request to the first network function entity, the second model update request including subscription association information and the local model parameters. The AF entity can extract the required information from the first model update request to generate a second model update request that can be sent to the NWDAF.

[0411] Furthermore, the AF entity can also receive a notification from the first network function entity that the model update is complete. The notification includes model information of the updated global machine learning model and may also include subscription association information. The model information includes information about the second model file. Optionally, the AF entity can obtain the second model file based on the information of the second model file.

[0412] In another optional embodiment, the AF entity is an untrusted AF entity; the AF entity can send a third model update request to the NEF entity, wherein the third model update request includes at least one of subscription association information, UE identifier, application identifier, and first network function entity identifier, as well as the local model parameters, so that the NEF entity determines the first network function entity based on the third model update request, and sends a fourth model update request to the first network function entity, wherein the fourth model update request may include local model parameters, and may also include subscription association information, etc. The AF entity can extract the required information from the first model update request to generate a third model update request that can be sent to the NEF; the NEF entity can extract the required information from the third model update request to generate a fourth model update request that can be sent to the first network function entity.

[0413] Furthermore, the AF entity can also receive a notification from the NEF entity that the model update is complete. The notification includes model information of the global machine learning model updated by the first network function entity, and may also include subscription association information, etc. The model information includes information of the second model file. Optionally, the AF entity can obtain the second model file based on the information of the second model file.

[0414] Based on the above embodiments, the AF entity can send a model update response to the UE through the application layer. The model update response includes information of the second model file and may also include subscription association information, UE identifier, etc.

[0415] In one or more possible embodiments, before S901 receives the model update information carrying the local model parameters sent by the UE through the application layer, it may further include:

[0416] Receive model acquisition information carrying UE identifier and / or model description information sent by the UE through the application layer;

[0417] The model acquisition information is sent to the first network function entity based on the UE identifier and / or model description information;

[0418] Information about the first model file of the global machine learning model is obtained from the first network functional entity and sent to the UE through the application layer.

[0419] The model description information includes at least one of the following: application identifier, application sub-feature identifier, time information, location information, and other model feature information.

[0420] Based on the above embodiments, the step of sending the model acquisition information to the first network function entity according to the UE identifier and / or model description information includes:

[0421] If the AF entity is a trusted AF entity, then select the first network function entity serving the UE, and / or the first network function entity capable of providing a global machine learning model that satisfies the model description information, and send the model acquisition information to the first network function entity; or...

[0422] If the AF entity is an untrusted AF entity, the model acquisition information is sent to the NEF entity. The NEF entity then selects a first network function entity serving the UE and / or a first network function entity capable of providing a global machine learning model that satisfies the model description information, and sends the model acquisition information to the first network function entity. In this embodiment, optionally, the AF entity is a trusted AF entity. The AF entity can select, based on the first model acquisition request, a candidate first network function entity serving the UE and / or a candidate first network function entity capable of providing a global machine learning model that satisfies the model description information, and determine it as the first network function entity. Then, it sends a second model acquisition request to the first network function entity, wherein the second model acquisition request may include the UE identifier and / or the model description information, so that the first network function entity determines the global machine learning model and generates subscription association information for the UE. The AF entity can extract the required information from the first model acquisition request and generate a second model acquisition request that can be sent to the NWDAF. Finally, it receives a model acquisition notification from the first network function entity, wherein the model acquisition notification includes model information and subscription association information of the global machine learning model, and the model information includes information from the first model file.

[0423] Optionally, the first model file can be obtained based on the information in the first model file.

[0424] Optionally, the AF entity may also be an untrusted AF entity; the AF entity may send a third model acquisition request to the NEF entity according to the first model acquisition request, wherein the third model acquisition request includes the UE identifier and / or the model description information, so that the NEF entity selects a candidate first network function entity serving the UE, and / or a candidate first network function entity that can provide a global machine learning model that satisfies the model description information, determines it as the first network function entity, and sends a fourth model acquisition request to the first network function entity, wherein the fourth model acquisition request includes the UE identifier and / or the model description information, so that the first network function entity determines the global machine learning model and generates subscription association information for the UE, wherein the AF entity may extract the required information from the first model acquisition request and generate a third model acquisition request that can be sent to the NEF, and the NEF entity may extract the required information from the third model acquisition request and generate a fourth model acquisition request that can be sent to the NWDAF; receive a model acquisition notification sent by the NEF entity, wherein the model acquisition notification includes model information and subscription association information of the global machine learning model provided by the first network function entity, and the model information includes information of the first model file.

[0425] Optionally, the first model file can be obtained based on the information in the first model file.

[0426] Based on the above embodiments, the step of obtaining information about the first model file of the global machine learning model from the first network functional entity and sending it to the UE through the application layer includes:

[0427] The information of the first model file of the global machine learning model can be obtained directly or through the NEF entity from the first network functional entity.

[0428] A model acquisition response is sent to the UE via the application layer, wherein the model acquisition response includes information about the first model file, used to acquire the first model file based on the information in the first model file. Optionally, the model acquisition response may also include subscription-related information and UE identifiers, etc. In one or more possible embodiments, the method further includes:

[0429] Establish a PDU session between the UE and the AF entity to connect the UE and the AF entity at the application layer. The PDU session includes: a PDU session dedicated to transmitting machine learning model data, or a PDU session dedicated to carrying specific application data.

[0430] In one or more possible embodiments, the first network functional entity is an NWDAF entity.

[0431] Example 10

[0432] Figure 16 This is a flowchart of the machine learning model processing method provided in this embodiment. Figure 16 As shown, this embodiment provides a method in which the execution subject is a first network functional entity used to provide the model. The specific steps of this machine learning model processing method are as follows:

[0433] S1001. Obtain the local model parameters of the trained local machine learning model sent by the user equipment (UE), wherein the local machine learning model is obtained based on the global machine learning model trained by the first network function entity.

[0434] S1002. Update the global machine learning model according to the local model parameters.

[0435] Furthermore, after updating the global machine learning model based on the local model parameters, the process may further include:

[0436] S1003. Send the updated global model parameters of the global machine learning model to the UE. The global model parameters are used to update the trained local machine learning model.

[0437] The global model parameters are obtained by the first network function entity updating the global machine learning model using local model parameters sent by at least one UE.

[0438] Based on the above embodiments, before obtaining the local model parameters of the trained local machine learning model sent by the user equipment (UE), the method further includes:

[0439] The information of the first model file of the global machine learning model is sent to the UE so that the UE can create a local machine learning model based on the first model file.

[0440] As an example, the specific steps of the machine learning model processing method executed by the first network functional entity are as follows: The machine learning model processing method provided in this embodiment is the method on the first network functional entity side in the above embodiment. Its principle and technical effect can be found in the above embodiment, and will not be repeated here.

[0441] In one or more possible embodiments, obtaining the local model parameters of the trained local machine learning model sent by the user equipment (UE) includes:

[0442] Receive a second model update request sent directly by the AF entity or through the NEF entity, wherein the second model update request is sent by the AF entity after receiving the first model update request sent by the UE through the application layer, and the second model update request includes local model parameters, and may also include subscription association information, etc.; obtain the local model parameters from the second model update request; or

[0443] The system receives a fifth model update request sent by the AMF entity, wherein the fifth model update request is sent by the AMF entity after receiving the first NAS message from the UE requesting a model update, and the fifth model update request includes local model parameters, and may also include subscription association information, etc.; the system obtains the local model parameters from the fifth model update request.

[0444] In one or more possible embodiments, before obtaining the local model parameters of the trained local machine learning model sent by the user equipment (UE), the method further includes:

[0445] Receive the UE's request to obtain the global machine learning model;

[0446] Determine the global machine learning model based on the acquisition request;

[0447] The information of the first model file of the global machine learning model is sent to the UE, and the UE creates a local machine learning model based on the first model file.

[0448] In the case of model requests, subscription association information can also be generated for the UE. The subscription association information can be a subscription association ID, which can be assigned to different UEs subscribing to the same model, or different subscription association IDs can be assigned to different UEs subscribing to the same model.

[0449] In one or more possible embodiments, receiving the UE's request to obtain the global machine learning model may specifically include:

[0450] Receive a second model acquisition request sent directly by the AF entity or through the NEF entity, wherein the second model acquisition request is sent by the AF entity after receiving the first model acquisition request sent by the UE through the application layer, and the second model acquisition request includes UE identifier and / or model description information; or

[0451] The system receives a fifth model acquisition request from the AMF entity, wherein the fifth model acquisition request is sent by the AMF entity after receiving the third NAS message from the UE requesting to acquire a model, and the fifth model acquisition request includes the UE identifier and / or model description information.

[0452] Based on the above embodiments, sending the information of the first model file of the global machine learning model to the UE includes:

[0453] Send a model acquisition notification to the AF entity or AMF entity, or send a model acquisition notification to the AF entity through the NEF entity, wherein the model acquisition notification includes model information of the global machine learning model, and may also include subscription association information, etc., and the model information includes information of the first model file, so that the second functional entity sends the information of the first model file to the UE.

[0454] In one or more possible embodiments, the first network functional entity is an NWDAF entity.

[0455] Example 11

[0456] Figure 17 This is a structural diagram of a user equipment according to an embodiment of the present invention. The user equipment provided in this embodiment can execute the processing flow provided in the method embodiment on the user equipment side, such as... Figure 17 As shown, the user equipment 1100 includes a memory 1101, a transceiver 1102, and a processor 1103.

[0457] Among them, Figure 17 In this context, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits together, represented by one or more processors (processor 1103) and memory (memory 1101). The bus architecture can also link various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. The transceiver 1102 can be multiple elements, including transmitters and receivers, providing units for communicating with various other devices over transmission media, including wireless channels, wired channels, optical fibers, etc. The processor 1101 is responsible for managing the bus architecture and general processing, and the memory 1103 can store data used by the processor 1103 during operation.

[0458] The processor 1103 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor can also adopt a multi-core architecture.

[0459] Memory 1101 is used to store computer programs; transceiver 1102 is used to send and receive data under the control of processor 1103; processor 1103 is used to read the computer program in memory 1101 and perform the following operations:

[0460] Identify local training data relevant to the target application;

[0461] The local machine learning model of the target application is trained based on the local training data to obtain the local model parameters of the trained local machine learning model, wherein the local machine learning model is obtained based on the global machine learning model trained by the first network functional entity.

[0462] The local model parameters are sent to the first network function entity, and the local model parameters are used to update the global machine learning model.

[0463] In one or more possible embodiments, the processor 1103 is further configured to:

[0464] Obtain the updated global model parameters of the global machine learning model from the first network functional entity; the global model parameters are used to update the trained local machine learning model.

[0465] The global model parameters are obtained by the first network function entity updating the global machine learning model using local model parameters sent by at least one UE.

[0466] In one or more possible embodiments, when the processor 1103 sends the local model parameters to the first network functional entity, it is configured to:

[0467] The model update information, carrying the local model parameters, is sent from the application layer to the application function (AF) entity, so that the AF entity can then send the model update information to the first network function entity; or,

[0468] The model update information, carrying the local model parameters, is sent from the application layer to the application function (AF) entity, and the AF entity then sends the model update information to the first network function entity through the network capability exposure (NEF) entity; or...

[0469] The non-access stratum sends model update information carrying the local model parameters to the Access and Mobility Management Function (AMF) entity, so that the AMF entity can send the model update information to the first network function entity.

[0470] In one or more possible embodiments, the model update information may further include any one or more of the following: UE identifier, application identifier, and first network function entity identifier, for use by the AF entity or the AMF entity to determine the first network function entity and / or the global machine learning model.

[0471] In one or more possible embodiments, when the processor 1103 obtains the updated global model parameters of the global machine learning model from the first network functional entity, it is configured to:

[0472] Receive a model update response sent by an AF entity or an AMF entity, wherein the model update response includes information on the global model parameters of the updated global machine learning model.

[0473] In one or more possible embodiments, when the processor 1103 obtains a local machine learning model based on the global machine learning model trained on the first network functional entity, it is configured to:

[0474] Obtain the first model file of the global machine learning model from the first network functional entity;

[0475] Create a local machine learning model based on the first model file.

[0476] In one or more possible embodiments, when the processor 1103 obtains the first model file of the global machine learning model from the first network functional entity, it is configured to:

[0477] Send model acquisition information carrying UE identifier and / or model description information to the first network function entity;

[0478] Based on the information obtained from the model, information about the first model file of the global machine learning model is obtained from the first network functional entity.

[0479] In one or more possible embodiments, when the processor 1103 sends model acquisition information carrying UE identifier and / or model description information to the first network function entity, it is configured to:

[0480] Model acquisition information carrying UE identifier and / or model description information is sent to the AF entity through the application layer, so that the AF entity can send the model acquisition information to the first network function entity based on the UE identifier and / or model description information; or...

[0481] The application layer sends model acquisition information carrying the UE identifier and / or model description information to the AF entity, and the AF entity then sends the model acquisition information to the first network function entity through the NEF entity based on the UE identifier and / or model description information; or...

[0482] The non-access stratum sends model acquisition information carrying UE identifier and / or model description information to the AMF entity, so that the AMF entity can send the model acquisition information to the first network function entity according to the UE identifier and / or model description information.

[0483] In one or more possible embodiments, when the processor 1103 obtains the first model file of the global machine learning model from the first network functional entity, it is configured to:

[0484] Receive a model acquisition response sent by an AF entity or an AMF entity, wherein the model acquisition response includes information about a first model file, and is used to acquire the first model file based on the information about the first model file.

[0485] In one or more possible embodiments, the model description information includes at least one of the following:

[0486] Application identifier, application sub-feature identifier, time information, location information, and other model feature information.

[0487] The user equipment provided in this embodiment of the invention can be specifically used to execute the above-described user equipment-side method embodiments, and the specific functions will not be described in detail here.

[0488] Example 12

[0489] Figure 18 This is a structural diagram of the AF entity according to an embodiment of the present invention. The AF entity provided in this embodiment can execute the processing flow provided in the method embodiment on the AF entity side, such as... Figure 18 As shown, the AF entity 1200 includes a memory 1201, a transceiver 1202, and a processor 1203;

[0490] Among them, Figure 18 In this context, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors 1203 (represented by processor 1203) and memory 1201 (represented by memory 1201). The bus architecture can also link various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. The transceiver 1202 can be multiple elements, including transmitters and receivers, providing a unit for communicating with various other devices over transmission media, including wireless channels, wired channels, optical fibers, etc. Processors 1203 and 1201 are responsible for managing the bus architecture and general processing, and memory 1203 can store data used by processors 1203 during operation.

[0491] The processor 1203 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD), and the processor 1203 may also adopt a multi-core architecture.

[0492] Memory 1201 is used to store computer programs; transceiver 1202 is used to send and receive data under the control of processor 1203; processor 1203 is used to read the computer program in memory 1201 and perform the following operations:

[0493] Receive model update information sent by the UE through the application layer, carrying the local model parameters, wherein the local machine learning model is a global machine learning model provided to the UE in advance by the first network function entity;

[0494] The model update information is sent to the first network function entity, and the local model parameters are used to update the global machine learning model.

[0495] In one or more possible embodiments, processor 1203 is further configured to:

[0496] Information on the global model parameters of the updated global machine learning model is obtained from the first network functional entity. The global model parameters are used to update the trained local machine learning model.

[0497] The application layer sends a model update response to the UE, wherein the model update response includes information about the global model parameters;

[0498] The global model parameters are obtained by the first network function entity updating the global machine learning model using local model parameters sent by at least one UE.

[0499] In one or more possible embodiments, when the processor 1203 sends the model update information to the first network functional entity, it is configured to:

[0500] If the AF entity is a trusted AF entity, then the model update information is directly sent to the first network function entity; or

[0501] If the AF entity is an untrusted AF entity, the model update information is sent to the first network function entity through the NEF entity.

[0502] In one or more possible embodiments, when the processor 1203 obtains information about the global model parameters of the updated global machine learning model from the first network functional entity, it is configured to:

[0503] The updated global model parameters of the global machine learning model can be obtained directly or through the NEF entity from the first network functional entity.

[0504] In one or more possible embodiments, before receiving model update information carrying the local model parameters sent by the UE through the application layer, the processor 1203 is further configured to:

[0505] Receive model acquisition information carrying UE identifier and / or model description information sent by the UE through the application layer;

[0506] The model acquisition information is sent to the first network function entity based on the UE identifier and / or model description information;

[0507] Information about the first model file of the global machine learning model is obtained from the first network functional entity and sent to the UE through the application layer.

[0508] In one or more possible embodiments, when the processor 1203 sends the model acquisition information to the first network function entity based on the UE identifier and / or model description information, it is configured to:

[0509] If the AF entity is a trusted AF entity, then select the first network function entity serving the UE, and / or the first network function entity capable of providing a global machine learning model that satisfies the model description information, and send the model acquisition information to the first network function entity; or...

[0510] If the AF entity is an untrusted AF entity, the model acquisition information is sent to the NEF entity. The NEF entity then selects a first network function entity that serves the UE and / or a first network function entity that can provide a global machine learning model that satisfies the model description information, and sends the model acquisition information to the first network function entity.

[0511] In one or more possible embodiments, when the processor 1203 obtains information about the first model file of the global machine learning model from the first network functional entity and sends it to the UE through the application layer, it is configured to:

[0512] The information of the first model file of the global machine learning model can be obtained directly or through the NEF entity from the first network functional entity.

[0513] The application layer sends a model acquisition response to the UE, wherein the model acquisition response includes information about the first model file, and is used to acquire the first model file based on the information about the first model file.

[0514] The AF entity provided in this embodiment of the invention can be specifically used to execute the method embodiment on the AF entity side described above, and the specific functions will not be repeated here.

[0515] Example 13

[0516] Figure 19 This is a structural diagram of the first network functional entity used to provide a model according to an embodiment of the present invention. The first network functional entity provided in this embodiment can execute the processing flow provided in the method embodiment for providing the first network functional entity of the model, such as... Figure 19 As shown, the first network functional entity 1300 for providing the model includes a memory 1301, a transceiver 1302, and a processor 1303;

[0517] Among them, Figure 19 In this context, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors 1303 (represented by processor 1303) and memory 1301 (represented by memory 1301). The bus architecture can also link various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. The transceiver 1302 can be multiple elements, including transmitters and receivers, providing a unit for communicating with various other devices over transmission media, including wireless channels, wired channels, optical fibers, etc. Processors 1303 and 1301 are responsible for managing the bus architecture and general processing, and memory 1303 can store data used by processors 1303 during operation.

[0518] The processor 1303 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor 1303 can also adopt a multi-core architecture.

[0519] Memory 1301 is used to store computer programs; transceiver 1302 is used to send and receive data under the control of processor 1303; processor 1303 is used to read the computer program in memory 1301 and perform the following operations:

[0520] Obtain the local model parameters of the trained local machine learning model sent by the user equipment (UE), wherein the local machine learning model is obtained based on the global machine learning model trained by the first network function entity.

[0521] The global machine learning model is updated based on the local model parameters.

[0522] In one or more possible embodiments, after updating the global machine learning model according to the local model parameters, the processor 1303 is further configured to:

[0523] The updated global model parameters of the global machine learning model are sent to the UE, and the global model parameters are used to update the trained local machine learning model;

[0524] The global model parameters are obtained by the first network function entity updating the global machine learning model using local model parameters sent by at least one UE.

[0525] In one or more possible embodiments, before acquiring the local model parameters of the trained local machine learning model sent by the user equipment (UE), the processor 1303 is further configured to:

[0526] The information of the first model file of the global machine learning model is sent to the UE so that the UE can create a local machine learning model based on the first model file.

[0527] The first network functional entity for providing the model provided in this embodiment of the invention can be specifically used to execute the method embodiment on the first network functional entity side described above. The specific functions will not be repeated here.

[0528] Example 14

[0529] Figure 20 This is a structural diagram of a machine learning model processing apparatus provided in an embodiment of the present invention. The machine learning model processing apparatus provided in this embodiment can execute the processing flow provided in the method embodiment on the UE side, such as... Figure 20 As shown, the machine learning model processing device 1400 includes an acquisition unit 1401, a training unit 1402, and a transmission unit 1403.

[0530] Acquisition unit 1401 is used to determine local training data related to the target application;

[0531] Training unit 1402 is used to train the local machine learning model of the target application based on the local training data to obtain the local model parameters of the trained local machine learning model, wherein the local machine learning model is obtained based on the global machine learning model trained by the first network functional entity.

[0532] The sending unit 1403 is used to send the local model parameters to the first network function entity, and the local model parameters are used to update the global machine learning model.

[0533] In one or more possible embodiments, the acquisition unit 1401 is further configured to acquire the updated global model parameters of the global machine learning model from the first network functional entity;

[0534] The training unit 1402 is also used to update the trained local machine learning model according to the global model parameters;

[0535] The global model parameters are obtained by the first network function entity updating the global machine learning model using local model parameters sent by at least one UE.

[0536] In one or more possible embodiments, the sending unit 1403, when sending the local model parameters to the first network functional entity, is configured to:

[0537] The model update information, carrying the local model parameters, is sent from the application layer to the application function (AF) entity, so that the AF entity can then send the model update information to the first network function entity; or,

[0538] The model update information, carrying the local model parameters, is sent from the application layer to the application function (AF) entity, and the AF entity then sends the model update information to the first network function entity through the network capability exposure (NEF) entity; or...

[0539] The non-access stratum sends model update information carrying the local model parameters to the Access and Mobility Management Function (AMF) entity, so that the AMF entity can send the model update information to the first network function entity.

[0540] In one or more possible embodiments, the model update information may further include any one or more of the following: UE identifier, application identifier, and first network function entity identifier, for use by the AF entity or the AMF entity to determine the first network function entity and / or the global machine learning model.

[0541] In one or more possible embodiments, when the acquisition unit 1401 acquires the updated global model parameters of the global machine learning model from the first network functional entity, it is used to:

[0542] Receive a model update response sent by an AF entity or an AMF entity, wherein the model update response includes information on the global model parameters of the updated global machine learning model.

[0543] In one or more possible embodiments, the acquisition unit 1401 is further configured to acquire a first model file of the global machine learning model from the first network functional entity;

[0544] The training unit 1402 is also used to create a local machine learning model based on the first model file.

[0545] In one or more possible embodiments, when the acquisition unit 1401 acquires the first model file of the global machine learning model from the first network functional entity, it is used to:

[0546] The model acquisition information carrying the UE identifier and / or model description information is sent to the first network function entity through the sending unit 1403;

[0547] Based on the information obtained from the model, information about the first model file of the global machine learning model is obtained from the first network functional entity.

[0548] In one or more possible embodiments, when the sending unit 1403 sends model acquisition information carrying UE identifier and / or model description information to the first network function entity, it is used to:

[0549] Model acquisition information carrying UE identifier and / or model description information is sent to the AF entity through the application layer, so that the AF entity can send the model acquisition information to the first network function entity based on the UE identifier and / or model description information; or...

[0550] The application layer sends model acquisition information carrying the UE identifier and / or model description information to the AF entity, and the AF entity then sends the model acquisition information to the first network function entity through the NEF entity based on the UE identifier and / or model description information; or...

[0551] The non-access stratum sends model acquisition information carrying UE identifier and / or model description information to the AMF entity, so that the AMF entity can send the model acquisition information to the first network function entity according to the UE identifier and / or model description information.

[0552] In one or more possible embodiments, the acquisition unit 1401 acquires a first model file of the global machine learning model from the first network functional entity, including:

[0553] Receive a model acquisition response sent by an AF entity or an AMF entity, wherein the model acquisition response includes information about a first model file, and is used to acquire the first model file based on the information about the first model file.

[0554] In one or more possible embodiments, the model description information includes at least one of the following:

[0555] Application identifier, application sub-feature identifier, time information, location information, and other model feature information.

[0556] The machine learning model processing device provided in this embodiment of the invention can be specifically used to execute the above-described method embodiment on the UE side; its specific functions will not be elaborated here.

[0557] Example 15

[0558] Figure 21 This is a structural diagram of a machine learning model processing apparatus provided in an embodiment of the present invention. The machine learning model processing apparatus provided in this embodiment can execute the processing flow provided in the method embodiment for the AF entity side, such as... Figure 21 As shown, the machine learning model processing device 1500 includes: a receiving unit 1501, a sending unit 1502, and an acquisition unit 1503.

[0559] The receiving unit 1501 is used to receive model update information carrying the local model parameters sent by the UE through the application layer, wherein the local machine learning model is a global machine learning model provided to the UE in advance by the first network function entity.

[0560] The sending unit 1502 is used to send the model update information to the first network function entity, and the local model parameters are used to update the global machine learning model.

[0561] In one or more possible embodiments, it also includes:

[0562] The acquisition unit 1503 is used to acquire information on the global model parameters of the updated global machine learning model from the first network functional entity, wherein the global model parameters are used to update the trained local machine learning model.

[0563] The sending unit 1502 is further configured to send a model update response to the UE through the application layer, wherein the model update response includes information about the global model parameters;

[0564] The global model parameters are obtained by the first network function entity updating the global machine learning model using local model parameters sent by at least one UE.

[0565] In one or more possible embodiments, when sending the model update information to the first network functional entity, the sending unit 1502 is configured to:

[0566] If the AF entity is a trusted AF entity, then the model update information is directly sent to the first network function entity; or

[0567] If the AF entity is an untrusted AF entity, the model update information is sent to the first network function entity through the NEF entity.

[0568] In one or more possible embodiments, when the acquisition unit 1503 acquires information about the global model parameters of the updated global machine learning model from the first network functional entity, it is used to:

[0569] The updated global model parameters of the global machine learning model can be obtained directly or through the NEF entity from the first network functional entity.

[0570] In one or more possible embodiments, before receiving the model update information carrying the local model parameters sent by the UE through the application layer, the method further includes:

[0571] The receiving unit 1501 is further configured to receive model acquisition information carrying UE identifier and / or model description information sent by the UE through the application layer;

[0572] The sending unit 1502 is further configured to send the model acquisition information to the first network function entity according to the UE identifier and / or model description information;

[0573] The acquisition unit 1503 is further configured to acquire information about the first model file of the global machine learning model from the first network functional entity;

[0574] The transmitting unit 1502 is further configured to transmit the data to the UE via the application layer. In one or more possible embodiments,

[0575] The machine learning model processing device provided in this embodiment of the invention can be specifically used to execute the above-described method embodiment on the AF entity side; its specific functions will not be elaborated here.

[0576] Example 16

[0577] Figure 22 This is a structural diagram of a machine learning model processing apparatus provided in an embodiment of the present invention. The machine learning model processing apparatus provided in this embodiment can execute the processing flow provided in the method embodiment for providing the first network functional entity side of the model, such as... Figure 22 As shown, the machine learning model processing device 1600 includes: an acquisition unit 1601, a model update unit 1602, and a sending unit 1603.

[0578] The acquisition unit 1601 is used to acquire the local model parameters of the trained local machine learning model sent by the user equipment (UE), wherein the local machine learning model is obtained based on the global machine learning model trained by the first network function entity.

[0579] The model update unit 1602 is used to update the global machine learning model according to the local model parameters.

[0580] In one or more possible embodiments, it also includes:

[0581] The sending unit 1603 is used to send the updated global model parameters of the global machine learning model to the UE, wherein the global model parameters are used to update the trained local machine learning model.

[0582] The global model parameters are obtained by the first network function entity updating the global machine learning model using local model parameters sent by at least one UE.

[0583] In one or more possible embodiments, before acquiring the local model parameters of the trained local machine learning model sent by the user equipment (UE), the sending unit 1603 is further configured to:

[0584] The information of the first model file of the global machine learning model is sent to the UE so that the UE can create a local machine learning model based on the first model file.

[0585] The machine learning model processing apparatus provided in this embodiment of the invention can be specifically used to execute the method embodiment on the first network functional entity side described above; the specific functions will not be elaborated here.

[0586] It should be noted that the division of units in the above embodiments of this application is illustrative and only represents a logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units.

[0587] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0588] Example 17

[0589] Embodiment 11 of this application provides a computer-readable storage medium storing a computer program for causing a processor to execute a machine learning model processing method on the UE side.

[0590] Example 18

[0591] Embodiment 11 of this application provides a computer-readable storage medium storing a computer program for causing a processor to execute a machine learning model processing method on the AF entity side.

[0592] Example 19

[0593] Embodiment 11 of this application provides a computer-readable storage medium storing a computer program for causing a processor to execute a machine learning model processing method for providing a first network functional entity side of a model.

[0594] The computer-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic storage (e.g., floppy disks, hard disks, magnetic tapes, magneto-optical disks (MOs), etc.), optical storage (e.g., CDs, DVDs, BDs, HVDs, etc.), and semiconductor storage (e.g., ROMs, EPROMs, EEPROMs, non-volatile memory (NAND flash), solid-state drives (SSDs)).

[0595] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0596] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-executable instructions. These computer-executable instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1A device that provides the functions specified in one or more boxes.

[0597] These processor-executable instructions may also be stored in a processor-readable memory that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the processor-readable memory produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0598] These processors can execute instructions that can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0599] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A machine learning model processing method, characterized by, The method is applied to a user equipment (UE), and comprises the following steps: determining local training data related to a target application; training a local machine learning model of the target application according to the local training data to obtain local model parameters of the trained local machine learning model, wherein the local machine learning model is obtained according to a global machine learning model trained by a first network function entity; sending the local model parameters to the first network function entity, wherein the local model parameters are used to update the global machine learning model; the step of sending the local model parameters to the first network function entity comprises: sending model update information carrying the local model parameters to an application function (AF) entity through an application layer, so as to send the model update information to the first network function entity through the AF entity, wherein the AF entity is a trusted entity; or sending model update information carrying the local model parameters to an application function (AF) entity through an application layer, so as to send the model update information to the first network function entity through the AF entity through a network exposure function (NEF) entity, wherein the AF entity is an untrusted entity; or sending model update information carrying the local model parameters to an access and mobility management function (AMF) entity through a non-access layer, so as to send the model update information to the first network function entity through the AMF entity.

2. The method of claim 1, wherein, The method further comprises the following steps: obtaining global model parameters of an updated global machine learning model from the first network function entity, wherein the global model parameters are used to update the trained local machine learning model; wherein the global model parameters are obtained by the first network function entity by updating the global machine learning model using local model parameters sent by at least one UE.

3. The method of claim 1, wherein, The model update information further comprises any one or more of the following: a UE identifier, an application identifier, and a first network function entity identifier, so as to enable the AF entity or the AMF entity to determine the first network function entity and / or the global machine learning model.

4. The method of claim 2, wherein, The step of obtaining global model parameters of an updated global machine learning model from the first network function entity comprises the following steps: receiving a model update response sent by the AF entity or the AMF entity, wherein the model update response comprises information of the global model parameters of the updated global machine learning model.

5. The method of claim 1, wherein, The local machine learning model is obtained according to a global machine learning model trained by a first network function entity, and comprises the following steps: obtaining a first model file of the global machine learning model from the first network function entity; creating a local machine learning model according to the first model file.

6. The method of claim 5, wherein, The step of obtaining a first model file of the global machine learning model from the first network function entity comprises the following steps: sending model obtaining information carrying a UE identifier and / or model description information to the first network function entity; obtaining information of the first model file of the global machine learning model from the first network function entity based on the model obtaining information.

7. The method of claim 6, wherein, The step of sending model obtaining information carrying a UE identifier and / or model description information to the first network function entity comprises the following steps: sending, by an application layer, model obtaining information carrying UE identity and / or model description information to an AF entity, so as to send, by the AF entity, the model obtaining information to the first network function entity according to the UE identity and / or the model description information; or sending, by an application layer, model obtaining information carrying UE identity and / or model description information to an AF entity, so as to send, by the AF entity, the model obtaining information to the first network function entity according to the UE identity and / or the model description information through a NEF entity; or sending, by a non-access layer, model obtaining information carrying UE identity and / or model description information to an AMF entity, so as to send, by the AMF entity, the model obtaining information to the first network function entity according to the UE identity and / or the model description information.

8. The method according to claim 6 or 7, characterized in that, The first model file of the global machine learning model obtained from the first network function entity comprises: receiving model obtaining response sent by the AF entity or the AMF entity, wherein the model obtaining response comprises information of the first model file, and the information of the first model file is used to obtain the first model file.

9. The method according to claim 6 or 7, characterized in that, The model description information comprises at least one of the following: application identity, application sub-characteristic identity, time information, location information, and other model characteristic information. 10.A method for processing a machine learning model, the method comprising: The method applied to an application function (AF) entity comprises: receiving model update information carrying local model parameters of a trained local machine learning model sent by a UE through an application layer, wherein the local machine learning model is a global machine learning model of a first network function entity provided to the UE in advance; sending the model update information to the first network function entity, wherein the local model parameters are used to update the global machine learning model; The method applied to an application function (AF) entity comprises: if the AF entity is a trusted AF entity, directly sending the model update information to the first network function entity; or if the AF entity is an untrusted AF entity, sending the model update information to the first network function entity through a NEF entity.

11. The method of claim 10, wherein, Further comprising: obtaining information of global model parameters of an updated global machine learning model from the first network function entity, wherein the global model parameters are used to update the trained local machine learning model; sending model update response to the UE through an application layer, wherein the model update response comprises information of the global model parameters; The global model parameters are obtained by the first network function entity by updating the global machine learning model using at least one local model parameter sent by a UE.

12. The method of claim 11, wherein, The method applied to an application function (AF) entity comprises: directly or through a NEF entity, obtaining information of global model parameters of an updated global machine learning model from the first network function entity.

13. The method of claim 10, wherein, Before the method applied to an application function (AF) entity comprises: receiving model obtaining information carrying UE identification and / or model description information sent by the UE through an application layer; sending the model obtaining information to the first network function entity according to the UE identification and / or model description information; obtaining information of a first model file of a global machine learning model from the first network function entity and sending the information to the UE through the application layer.

14. The method of claim 13, wherein, The sending of the model obtaining information to the first network function entity according to the UE identification and / or model description information comprises: if the AF entity is a trusted AF entity, selecting a first network function entity serving the UE and / or a first network function entity capable of providing a global machine learning model satisfying the model description information, and sending the model obtaining information to the first network function entity; or if the AF entity is an untrusted AF entity, sending the model obtaining information to a NEF entity, selecting a first network function entity serving the UE and / or a first network function entity capable of providing a global machine learning model satisfying the model description information through the NEF entity, and sending the model obtaining information to the first network function entity.

15. The method according to claim 13 or 14, characterized in that, The obtaining of the information of the first model file of the global machine learning model from the first network function entity and the sending of the information to the UE through the application layer comprise: obtaining the information of the first model file of the global machine learning model from the first network function entity directly or through the NEF entity; sending a model obtaining response to the UE through the application layer, wherein the model obtaining response comprises the information of the first model file, and the information of the first model file is used to obtain the first model file. 16.A machine learning model processing method, comprising: The method is applied to a first network function entity for providing a model, and the method comprises: obtaining local model parameters of a trained local machine learning model sent by a user equipment (UE), wherein the local machine learning model is obtained according to a global machine learning model trained by the first network function entity; wherein the obtaining of the local model parameters of the trained local machine learning model sent by the UE comprises: obtaining the local model parameters of the trained local machine learning model sent by the UE through an application function (AF) entity; wherein the AF entity is a trusted entity; or obtaining the local model parameters of the trained local machine learning model sent by the UE through an AF entity and a network capability exposure function (NEF) entity; wherein the AF entity is an untrusted entity; or obtaining the local model parameters of the trained local machine learning model sent by the UE through a mobility management function (AMF) entity; updating the global machine learning model according to the local model parameters.

17. The method of claim 16, wherein, After the updating of the global machine learning model according to the local model parameters, the method further comprises: sending global model parameters of the updated global machine learning model to the UE, wherein the global model parameters are used to update the trained local machine learning model. The global model parameter is obtained by the first network function entity updating the global machine learning model by using at least one local model parameter sent by a UE.

18. The method of claim 16, wherein, Before the acquiring of the local model parameter of the trained local machine learning model sent by the user equipment (UE), the method further comprises: sending information of a first model file of the global machine learning model to the UE, so that the UE creates a local machine learning model according to the first model file.

19. A user equipment, comprising: The method comprises the following steps of: a memory, a transceiver, and a processor: The memory is configured to store a computer program; the transceiver is configured to transceive data under the control of the processor; and the processor is configured to read the computer program in the memory and perform the following operations: determining local training data related to a target application; training a local machine learning model of the target application according to the local training data to obtain a local model parameter of a trained local machine learning model, wherein the local machine learning model is obtained according to a global machine learning model trained by a first network function entity; sending the local model parameter to the first network function entity, wherein the local model parameter is used to update the global machine learning model; and the sending of the local model parameter to the first network function entity comprises: sending model update information carrying the local model parameter to an application function (AF) entity through an application layer, so that the AF entity sends the model update information to the first network function entity; wherein the AF entity is a trusted entity; or sending model update information carrying the local model parameter to an application function (AF) entity through an application layer, so that the AF entity sends the model update information to the first network function entity through a network exposure function (NEF) entity; wherein the AF entity is an untrusted entity; or 20. An AF entity, comprising: sending model update information carrying the local model parameter to an access and mobility management function (AMF) entity through a non-access layer, so that the AMF entity sends the model update information to the first network function entity. The method comprises the following steps of: a memory, a transceiver, and a processor: The memory is configured to store a computer program; the transceiver is configured to transceive data under the control of the processor; and the processor is configured to read the computer program in the memory and perform the following operations: receiving model update information carrying a local model parameter of a trained local machine learning model sent by a user equipment (UE) through an application layer, wherein the local machine learning model is a global machine learning model provided by a first network function entity to the UE in advance; sending the model update information to the first network function entity, wherein the local model parameter is used to update the global machine learning model; when the processor sends the model update information to the first network function entity, it is configured to: if the AF entity is a trusted AF entity, directly send the model update information to the first network function entity; or if the AF entity is an untrusted AF entity, send the model update information to the first network function entity through a network exposure function (NEF) entity.

21. A first network functional entity for providing a model, characterized in that, Includes memory, transceiver, and processor: A memory for storing computer programs; a transceiver for sending and receiving data under the control of the processor; and a processor for reading the computer programs from the memory and performing the following operations: Obtain the local model parameters of the trained local machine learning model sent by the user equipment (UE), wherein the local machine learning model is obtained based on the global machine learning model trained by the first network function entity; wherein obtaining the local model parameters of the trained local machine learning model sent by the user equipment (UE) includes: obtaining the local model parameters of the trained local machine learning model sent by the user equipment (UE) through an application function (AF) entity; wherein the AF entity is a trusted entity; or, The local model parameters of the trained local machine learning model sent by the user equipment UE are obtained through the application function AF entity and the network capability open function NEF entity; wherein the AF entity is an untrusted entity; or, the local model parameters of the trained local machine learning model sent by the user equipment UE are obtained through the mobility management function AMF entity. The global machine learning model is updated based on the local model parameters.

22. A machine learning model processing apparatus, comprising: Applied to a UE, the device includes: The acquisition unit is used to determine the local training data relevant to the target application; The training unit is used to train the local machine learning model of the target application based on the local training data to obtain the local model parameters of the trained local machine learning model, wherein the local machine learning model is obtained based on the global machine learning model trained by the first network functional entity. A sending unit is configured to send the local model parameters to the first network function entity, wherein the local model parameters are used to update the global machine learning model; The sending unit is specifically configured to: send model update information carrying the local model parameters to the application function (AF) entity through the application layer, so that the model update information is sent to the first network function entity through the AF entity; wherein the AF entity is a trusted entity; or... The application layer sends model update information carrying the local model parameters to the application function (AF) entity, which then sends the model update information to the first network function entity via the network capability exposure (NEF) entity; wherein the AF entity is an untrusted entity; or, The model update information carrying the local model parameters is sent to the Access and Mobility Management Function (AMF) entity through the non-access stratum, so that the model update information can be sent to the first network function entity through the AMF entity.

23. A machine learning model processing apparatus, comprising: Applied to AF entities, the device includes: The receiving unit is used to receive model update information sent by the UE through the application layer, which carries local model parameters of the trained local machine learning model, wherein the local machine learning model is a global machine learning model provided to the UE in advance by the first network function entity. A sending unit is used to send the model update information to the first network function entity, wherein the local model parameters are used to update the global machine learning model; The sending unit is specifically configured to: if the AF entity is a trusted AF entity, directly send the model update information to the first network function entity; or if the AF entity is an untrusted AF entity, send the model update information to the first network function entity through an NEF entity.

24. A machine learning model processing apparatus, comprising: The device is applied to a first network function entity for providing a model, and the device comprises: A obtaining unit is configured to obtain local model parameters of a trained local machine learning model sent by a user equipment (UE), wherein the local machine learning model is obtained according to a global machine learning model trained by a first network function entity; wherein the obtaining of the local model parameters of the trained local machine learning model sent by the UE comprises: obtaining, by an application function (AF) entity, the local model parameters of the trained local machine learning model sent by the UE; wherein the AF entity is a trusted entity; or obtaining, by an application function (AF) entity and a network capability exposure function (NEF) entity, the local model parameters of the trained local machine learning model sent by the UE; wherein the AF entity is an untrusted entity; or obtaining, by a mobility management function (AMF) entity, the local model parameters of the trained local machine learning model sent by the UE; A model updating unit is configured to update the global machine learning model according to the local model parameters.

25. A processor-readable storage medium, comprising: The processor readable storage medium stores a computer program, and the computer program is used to make the processor execute the method in any one of claims 1-9, 10-15 or 16-18.

Images