WO2023185822A1 - 网元注册方法、模型确定方法、装置、网元、通信系统及存储介质 - Google Patents

网元注册方法、模型确定方法、装置、网元、通信系统及存储介质 Download PDF

Info

Publication number
WO2023185822A1
WO2023185822A1 PCT/CN2023/084355 CN2023084355W WO2023185822A1 WO 2023185822 A1 WO2023185822 A1 WO 2023185822A1 CN 2023084355 W CN2023084355 W CN 2023084355W WO 2023185822 A1 WO2023185822 A1 WO 2023185822A1
Authority
WO
WIPO (PCT)
Prior art keywords
network element
information
federated learning
target
learning training
Prior art date
Application number
PCT/CN2023/084355
Other languages
English (en)
French (fr)
Inventor
程思涵
崇卫微
Original Assignee
维沃移动通信有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 维沃移动通信有限公司 filed Critical 维沃移动通信有限公司
Publication of WO2023185822A1 publication Critical patent/WO2023185822A1/zh

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W8/00—Network data management
    • H04W8/02—Processing of mobility data, e.g. registration information at HLR [Home Location Register] or VLR [Visitor Location Register]; Transfer of mobility data, e.g. between HLR, VLR or external networks
    • H04W8/08—Mobility data transfer
    • H04W8/14—Mobility data transfer between corresponding nodes

Definitions

  • This application belongs to the field of communication technology, and specifically relates to a network element registration method, a model determination method, a device, a network element, a communication system and a storage medium.
  • Federated learning can allow multiple clients to cooperate with each other under the coordination of a central server to obtain a complete machine learning model.
  • federated learning there is no additional enhancement or differentiated treatment for artificial intelligence algorithms such as federated learning.
  • network elements, base stations, user equipment (UE) and other participants that can be trained by default can participate.
  • Federated learning such unfiltered federated learning object selection, may cause problems such as loss of federated learning objects and reduced learning efficiency due to various reasons (such as poor UE network signal, etc.). Therefore, how to enhance the network element registration process in federated learning is an urgent problem to be solved.
  • Embodiments of the present application provide a network element registration method, model determination method, device, network element, communication system and storage medium, which can solve the problem of how to enhance the network element registration process in federated learning.
  • a network element registration method includes: a first network element sends a registration request to a second network element.
  • the registration request is used to request to register the federated learning capability information of the first network element to the second network element.
  • Network element, the federated learning capability information of the first network element includes at least one of the following: type information of federated learning training supported by the first network element, time information of federated learning training supported by the first network element, information of the federated learning training possessed by the first network element. Metadata information.
  • a network element registration device which is applied to the first network element.
  • the network element registration device includes: a sending module.
  • a sending module configured to send a registration request to the second network element.
  • the registration request is used to request to register the federated learning capability information of the first network element to the second network element.
  • the federated learning capability information of the first network element includes at least one of the following: Items: type information of federated learning training supported by the first network element, time information of federated learning training supported by the first network element, and metadata information possessed by the first network element.
  • a network element registration method includes: the second network element receives a registration request sent by the first network element, and the registration request is used to request to register the federated learning capability information of the first network element to the third network element.
  • the federated learning capability information of the first network element includes at least one of the following: the type information of federated learning training supported by the first network element, the time information of the federated learning training supported by the first network element, and the information of the federated learning training supported by the first network element. metadata information.
  • a network element registration device which is applied to the second network element.
  • the network element registration device includes: a receiving module.
  • the receiving module is configured to receive a registration request sent by the first network element.
  • the registration request is used to request to register the federated learning capability information of the first network element to the second network element.
  • the federated learning capability information of the first network element includes at least the following: One item: type information of federated learning training supported by the first network element, time information of federated learning training supported by the first network element, and metadata information possessed by the first network element.
  • a method for determining a model includes: a third network element sends a search request to the second network element.
  • the search request is used to request to find a network element that can perform federated learning training.
  • the search request includes the third network element.
  • Information, the first information includes at least one of the following: data analysis identification information corresponding to the target task, type information of federated learning training corresponding to the target task, time information of federated learning training corresponding to the target task, federated learning corresponding to the target task.
  • the network element type requirement information is used to indicate the network element type corresponding to the network element that needs to be found and can perform federated learning training.
  • the network element type includes the federated learning server network element type and/or the federated learning client network element type.
  • the network element quantity requirement information is used to indicate the quantity requirement of the required network elements that can perform federated learning training.
  • a model determination device includes: a sending module.
  • a sending module configured to send a search request to the second network element.
  • the search request is used to request to find a network element that can perform federated learning training.
  • the search request includes first information, and the first information includes at least one of the following: target task. Corresponding data analysis identification information, type information of federated learning training corresponding to the target task, time information of federated learning training corresponding to the target task, metadata information of network elements corresponding to federated learning training corresponding to the target task, and network element type requirements Information and network element quantity requirement information.
  • the network element type requirement information is used to indicate the network element type corresponding to the network element that needs to be searched for and can perform federated learning training.
  • the network element type includes the federated learning server network element type and/or federation Learning client network element type, network element quantity requirement information is used to indicate that the required search can be carried out The number of network elements required for federated learning training.
  • a network element in a seventh aspect, includes a processor and a memory.
  • the memory stores programs or instructions that can be run on the processor.
  • the program or instructions are implemented when executed by the processor. The steps of the network element registration method described in the first aspect.
  • a network element including a processor and a communication interface, wherein the communication interface is used to send a registration request to the second network element, and the registration request is used to request the federated learning capability of the first network element.
  • the information is registered to the second network element.
  • the federated learning capability information of the first network element includes at least one of the following: type information of federated learning training supported by the first network element, time information of federated learning training supported by the first network element, information on the federated learning training supported by the first network element, Metadata information possessed by network elements.
  • a network element in a ninth aspect, includes a processor and a memory.
  • the memory stores programs or instructions that can be run on the processor.
  • the program or instructions are implemented when executed by the processor. The steps of the network element registration method described in the third aspect.
  • a network element including a processor and a communication interface, wherein the communication interface is used to receive a registration request sent by the first network element, and the registration request is used to request federated learning of the first network element. Capability information is registered to the second network element.
  • the federated learning capability information of the first network element includes at least one of the following: type information of federated learning training supported by the first network element, time information of federated learning training supported by the first network element, Metadata information owned by a network element.
  • a network element in an eleventh aspect, includes a processor and a memory.
  • the memory stores programs or instructions that can be run on the processor. When the program or instructions are executed by the processor, Implement the steps of the model determination method described in the fifth aspect.
  • a network element including a processor and a communication interface, wherein the communication interface is used to send a search request to a second network element, and the search request is used to request to find a network element that can perform federated learning training.
  • the search request includes first information, and the first information includes at least one of the following: data analysis identification information corresponding to the target task, type information of federated learning training corresponding to the target task, and time of federated learning training corresponding to the target task.
  • the network element type requirement information is used to indicate the required network element capable of federated learning training.
  • the network element type corresponding to the network element includes the federated learning server network element type and/or the federated learning client network element type.
  • the network element quantity requirement information is used to indicate the required network elements that can perform federated learning training. quantity requirements.
  • a communication system including: a network element registration device as described in the second aspect, a network element registration device as described in the fourth aspect, and a model determination device as described in the sixth aspect; or, It includes: the network element as described in the seventh aspect, the ninth aspect and the eleventh aspect; or includes: the network element as described in the eighth aspect, the tenth aspect and the twelfth aspect.
  • the network element may be used to perform the steps of the method described in the first aspect, the third aspect and the fifth aspect.
  • a readable storage medium is provided. Programs or instructions are stored on the readable storage medium. When the programs or instructions are executed by a processor, the steps of the method described in the first aspect are implemented, or the steps of the method are implemented. The steps of the method as described in the third aspect, or the steps of implementing the method as described in the fifth aspect.
  • a chip in a fifteenth aspect, includes a processor and a communication interface.
  • the communication interface is coupled to the processor.
  • the processor is used to run programs or instructions to implement the method described in the first aspect. method, or implement the method as described in the third aspect, or implement the method as described in the fifth aspect.
  • a computer program/program product is provided, the computer program/program product is stored in a storage medium, and the computer program/program product is executed by at least one processor to implement as described in the first aspect
  • the first network element can register the federated learning capability information of the first network element into the second network element through a registration request, so as to register the first network element into the second network element, thereby
  • This enables other network elements to find the first network element that can perform federated learning training through the second network element, so as to solve the problem of how the first network element registers the federated learning capability information to the second network element and can be found by other network elements. The problem.
  • the second network element can register the federated learning capability information of the first network element into the second network element by receiving the registration request sent by the first network element, so as to register the first network element into the second network element, thereby This enables other network elements to find the first network element that can perform federated learning training, so as to solve the problem of how the first network element registers the federated learning capability information to the second network element and can be found by other network elements.
  • the third network element can use the search request to search the second network element for network elements that can perform federated learning training. Since the search request includes the first information, the third network element can find the target network element that matches the first information. . Thus, federated learning training is performed with the found target network elements that can be federated learning training.
  • Figure 1 is a schematic architectural diagram of a wireless communication system provided by an embodiment of the present application.
  • Figure 2 is a schematic diagram of the principle of horizontal federated learning provided by an embodiment of the present application.
  • Figure 3 is a schematic diagram of a neural network provided by an embodiment of the present application.
  • Figure 4 is a schematic diagram of a neuron provided by an embodiment of the present application.
  • Figure 5 is one of the flow charts of a network element registration method provided by an embodiment of the present application.
  • Figure 6 is the second flow chart of a network element registration method provided by an embodiment of the present application.
  • Figure 7 is the third flow chart of a network element registration method provided by an embodiment of the present application.
  • Figure 8 is the fourth flow chart of a network element registration method provided by an embodiment of the present application.
  • Figure 9 is one of the flow charts of a model determination method provided by an embodiment of the present application.
  • Figure 10 is the second flow chart of a model determination method provided by the embodiment of the present application.
  • Figure 11 is the third flow chart of a model determination method provided by the embodiment of the present application.
  • Figure 12 is the fourth flowchart of a model determination method provided by the embodiment of the present application.
  • Figure 13 is a flow chart of a network element registration method and a model determination method provided by an embodiment of the present application
  • Figure 14 is one of the structural schematic diagrams of a network element registration device provided by an embodiment of the present application.
  • Figure 15 is the second structural schematic diagram of a network element registration device provided by an embodiment of the present application.
  • Figure 16 is a schematic structural diagram of a model determination device provided by an embodiment of the present application.
  • Figure 17 is a schematic diagram of the hardware structure of a communication device provided by an embodiment of the present application.
  • Figure 18 is a schematic diagram of the hardware structure of a network element provided by an embodiment of the present application.
  • first, second, etc. in the description and claims of this application are used to distinguish similar objects and are not used to describe a specific order or sequence. It is to be understood that the terms so used are interchangeable under appropriate circumstances so that the embodiments of the present application can be practiced in sequences other than those illustrated or described herein, and that "first" and “second” are distinguished objects It is usually one type, and the number of objects is not limited.
  • the first object can be one or multiple.
  • “and/or” in the description and claims indicates at least one of the connected objects, and the character “/" generally indicates that the related objects are in an "or” relationship.
  • LTE Long Term Evolution
  • LTE-Advanced, LTE-A Long Term Evolution
  • CDMA Code Division Multiple Access
  • TDMA Time Division Multiple Access
  • FDMA Frequency Division Multiple Access
  • OFDMA Orthogonal Frequency Division Multiple Access
  • SC-FDMA Single-carrier Frequency Division Multiple Access
  • system and “network” in the embodiments of this application are often used interchangeably, and the described technology can be used not only for the above-mentioned systems and radio technologies, but also for other systems and radio technologies.
  • NR New Radio
  • the following description describes a New Radio (NR) system for example purposes, and uses NR terminology in much of the following description, but these techniques can also be applied to applications other than NR system applications, such as 6th Generation , 6G) communication system.
  • NR New Radio
  • FIG. 1 shows a block diagram of a wireless communication system to which embodiments of the present application are applicable.
  • the wireless communication system includes a terminal 11 and a network side device 12.
  • the terminal 11 can be a mobile phone, a tablet computer (Tablet Personal Computer), a laptop computer (Laptop Computer), or a notebook computer, a personal digital assistant (Personal Digital Assistant, PDA), a handheld computer, a netbook, or a super mobile personal computer.
  • Tablet Personal Computer Tablet Personal Computer
  • laptop computer laptop computer
  • PDA Personal Digital Assistant
  • PDA Personal Digital Assistant
  • UMPC ultra-mobile personal computer
  • UMPC mobile Internet device
  • Mobile Internet Device MID
  • augmented reality augmented reality, AR
  • VR virtual reality
  • robots wearable devices
  • VUE vehicle-mounted equipment
  • PUE pedestrian terminal
  • smart home home equipment with wireless communication functions, such as refrigerators, TVs, washing machines or furniture, etc.
  • game consoles personal computers (personal computers, PC), teller machines or self-service Terminal devices
  • wearable devices include: smart watches, smart bracelets, smart headphones, smart glasses, smart jewelry (smart bracelets, smart bracelets, smart rings, smart necklaces, smart anklets, smart anklets, etc.), Smart wristbands, smart clothing, etc.
  • the network side equipment 12 may include access network equipment or core network equipment, where the access network equipment 12 may also be called wireless access network equipment, radio access network (Radio Access Network, RAN), radio access network function or Wireless access network unit.
  • the access network device 12 may include a base station, a WLAN access point or a WiFi node, etc.
  • the base station may be called a Node B, an evolved Node B (eNB), an access point, a Base Transceiver Station (BTS), a radio Base station, radio transceiver, Basic Service Set (BSS), Extended Service Set (ESS), Home B-Node, Home Evolved B-Node, Transmitting Receiving Point (TRP) or all
  • eNB evolved Node B
  • BTS Base Transceiver Station
  • BSS Basic Service Set
  • ESS Extended Service Set
  • Home B-Node Home Evolved B-Node
  • TRP Transmitting Receiving Point
  • Core network equipment may include but is not limited to at least one of the following: core network nodes, core network functions, mobility management entities (Mobility Management Entity, MME), access mobility management functions (Access and Mobility Management Function, AMF), session management functions (Session Management Function, SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Services Discovery function (Edge Application Server Discovery Function, EASDF), unified data management (Unified Data Management, UDM), unified data warehousing (Unified Data Repository, UDR), home subscriber server (Home Subscriber Server, HSS), centralized network configuration ( Centralized network configuration, CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (Local NEF, or L-NEF), Binding Support Function (Binding Support Function, BSF), application function (Application Function, AF), etc.
  • MME mobility management entities
  • AMF Access and Mobility Management Function
  • SMF Session Management Function
  • Server A is the coordinator, responsible for sending federated learning tasks, models and other information to other clients and members, and collecting information such as models, parameters, gradients, change rates and other feedback from clients. Update the initial model, and then issue the updated model;
  • Figure 2 shows the principle diagram of horizontal federated learning;
  • 1 is Sending encrypted gradients
  • AI Artificial Intelligence
  • AI models have a variety of algorithm implementations, such as neural networks, decision trees, support vector machines, Bayesian classifiers, etc. This application takes neural network as an example for explanation, but does not limit the specific type of AI module.
  • FIG. 3 a schematic diagram of a neural network.
  • An input layer, hidden layer, and output layer composed of these many neurons is a neural network.
  • the number of hidden layers and the number of neurons in each layer is the "network structure" of the neural network.
  • the neural network is composed of neurons, as shown in Figure 4, a schematic diagram of neurons.
  • a1, a2,...aK i.e., X1...Xn above
  • w is the weight (multiplicative coefficient)
  • b is the bias (additive coefficient)
  • ⁇ (z) is the activation function
  • z is the output value.
  • Common activation functions include Sigmoid, tanh, ReLU (Rectified Linear Unit, linear rectification function, modified linear unit), etc.
  • the parameter information of each neuron and the algorithm used are combined to form the "parameter information" of the entire network, which is also an important part of the AI model file.
  • an AI model refers to a file containing elements such as network structure and parameter information.
  • the trained AI model can be directly reused by its framework platform without repeated construction or learning, and can be directly judged, identified, etc. Intelligent functions.
  • NWDAF Network Data Analysis Function
  • AnLF AnLF
  • MTLF Network Data Analysis Function
  • NWDAF-MTLF coordinator
  • NWDAF-MTLF1 NWDAF-MTLF1
  • NWDAF-MTLF2 NWDAF-MTLF2
  • NWDAF-MTLF server A, coordinator
  • server A coordinator
  • NWDAF-MTLF1 (Database B1, participant 1) can collect data from Suzhou area and train AI model locally in Suzhou;
  • NWDAF-MTLF2 (Database B2, participant 2) can collect data from Xuzhou area and train AI model locally in Suzhou.
  • devices in different domains such as NWDAF network elements, base stations, and UEs that can participate in federated learning need to initiate a registration process with network elements responsible for capability storage such as NRF, indicating that they have capability information to participate in federated learning.
  • a federated learning initiator such as MTRL network element
  • the federated learning initiator requests information from network elements such as NRF to find a suitable federated learning client (such as NWDAF network element).
  • NWDAF network element federated learning initiator
  • the federated learning initiator MTLF network element sends the model and federated learning results to the AnLF network element and can point out the information of the participants participating in the federated learning.
  • network elements such as NWDAF can register federated learning capability information into the NRF network element through a registration request, so as to register network elements such as NWDAF into the NRF network element, so that other network elements can pass through the NRF network.
  • Element search finds network elements that can perform federated learning training to solve the problem of how network elements such as NWDAF register federated learning capability information to NRF network elements and can be found by other network elements.
  • the NRF network element can register the federated learning capability information of the NWDAF and other network elements into the NRF network element by receiving the registration request sent by the NWDAF and other network elements, so as to realize the registration of the NWDAF and other network elements into the NRF network element.
  • other network elements can find network elements that can perform federated learning training through NRF network elements, so as to solve the problem of how network elements such as NWDAF use federated learning capability information. The problem is that it is registered to the NRF network element and can be found by other network elements.
  • the MTRL network element can search the NRF network element for a target network element that can perform federated learning training through a search request. Since the search request includes the first information, the MTRL network element can find the target network element that matches the first information. Target network element. Thus, federated learning training is performed with the found target network elements that can be federated learning training.
  • FIG. 5 shows a flow chart of a network element registration method provided by an embodiment of the present application.
  • the network element registration method provided by the embodiment of the present application may include the following steps 201 and 202.
  • Step 201 The first network element sends a registration request to the second network element.
  • the above-mentioned registration request is used to request to register the federated learning capability information of the first network element to the second network element.
  • Step 202 The second network element receives the registration request sent by the first network element.
  • the federated learning capability information of the first network element includes at least one of the following: type information of federated learning training supported by the first network element, time information of federated learning training supported by the first network element, time information of the federated learning training supported by the first network element, The metadata information possessed by the element.
  • the above-mentioned first network element can be an NWDAF network element or a NWDAF containing MTLF network element
  • the second network element is a network element responsible for capability storage.
  • This network element can be an NRF network element or a UDM network element.
  • DCAF data collection application function
  • the above-mentioned first network element may send an Nnrf_NFManagement_NFRegister Register registration request to the second network element to request that the first network element be registered with the second network element.
  • the above-mentioned federated learning training type information supported by the first network element is used to indicate the type of AI model training algorithm supported by the first network element.
  • the type information of federated learning and training supported by the first network element includes at least one of the following: indication information of whether the first network element supports federated learning and training, horizontal federated learning and training type, vertical federation Learning training type, federated learning server capabilities, federated learning client capabilities.
  • federated learning server capability is used to indicate whether the first network element has the federated learning server capability, or is used to indicate whether the first network element supports serving as a federated learning server.
  • the first network element can serve as a federated learning server, which has the ability to aggregate local model training information provided by various federated learning clients to generate a global model (Aggregate model), or coordinate federated learning (Coordinate federated learning). ) process capabilities.
  • a federated learning server which has the ability to aggregate local model training information provided by various federated learning clients to generate a global model (Aggregate model), or coordinate federated learning (Coordinate federated learning). ) process capabilities.
  • federated learning client (clients) capability is used to indicate whether the first network element has the federated learning client capability, or is used to indicate whether the first network element supports serving as a federated learning client.
  • the first network element can serve as a federated learning client, which has the ability to participate in federated learning, perform local model training, and provide local model training information.
  • the above-mentioned time information of the first network element supporting federated learning training is used to indicate the federated learning training time supported by the first network element.
  • Federated learning training can achieve better performance within this time period, for example : 10pm to 6am the next morning.
  • the metadata information possessed by the first network element is used to indicate the data information that the first network element can cover, obtain, and provide.
  • the metadata information of the first network element includes at least one of the following: input data type, output data type, data amount, and data range.
  • the above input data type refers to the input data type of federated learning training
  • the above output data type refers to the output data type of federated learning training.
  • the above input data type is used to indicate the data type that can be collected by the first network element and used as input data for the model training process, or the data type that can be obtained by the first network element.
  • the above-mentioned data type refers to whether the data in the area has obvious characteristics, such as gathering in the morning and spreading out at night.
  • the above data type may be a data category, such as: UE location information, UE time information, network element load information, network status information, network element resource information, etc.
  • the above data range refers to the service range of the first network element.
  • the above-mentioned data range includes at least one of the following: the service area of the first network element, the area where the first network element can collect data, and the objects from which the first network element can collect data.
  • the data collectable by the first network element includes metadata information and training data used for model training.
  • the area in which the first network element can collect data may be any of the following: the first network element service area, a sub-area range under the first network element service area, a more fine-grained Objects that can collect data granularity (for example: a UE list).
  • the objects from which the first network element can collect data may include one or more specific network elements, or one or more UEs.
  • the above-mentioned registration request also includes at least one of the following: identification information of the first network element, data analysis identification information supported by the first network element, and model filter information supported by the first network element. , information about the network element type of the first network element.
  • the network element type includes the federated learning server network element type and/or the federated learning client network element type.
  • the data analysis identification information supported by the first network element may be user mobility trajectory (UE mobility) information.
  • UE mobility user mobility trajectory
  • the above data analysis identification information corresponds to the type information of federated learning training.
  • the above data analysis identification information corresponds to the federated learning capability information of the first network element.
  • the federated learning capability information of the first network element corresponds to the data analysis identification information.
  • the federated learning capability information of the first network element may correspond to a certain data analysis identification information, that is, It means that the first network element has the federated learning capability for the task identified by the data analysis.
  • mapping relationship between the data analysis identification information and the federated learning capability information of the first network element.
  • the mapping relationship can be one-to-one, many-to-one or one-to-many; wherein, in the mapping relationship When it is one-to-one, the federated learning capability information of the first network element also corresponds to different data analysis identification information.
  • the first network element supports 10 data analysis identification information
  • data analysis identification 1-3 corresponds to support horizontal federated learning training
  • data analysis identification 4-6 corresponds to supports vertical federated learning training
  • data analysis identification 7-10 corresponds to not supporting federation. Learn and train.
  • the identification information of the first network element may be any of the following: fully qualified domain name (Fully Qualified Domain Name, FQDN) information, IP address information.
  • fully qualified domain name Fully Qualified Domain Name, FQDN
  • IP address information IP address information.
  • the fully qualified domain name information is used to indicate the location of the first network element and to connect to the first network element.
  • Model filter information can also be called model effective range information, and the model effective range information is used to indicate the effective range of the model generated by the first network element training.
  • the above-mentioned model valid range information includes the slices to which the model is applicable, the external data network name (Data Network Name, DNN) to which the model is applicable, the area to which the model is applicable, the time when the model is applicable, etc.
  • the applicable time of the above model can be a large time range, such as within a year; or it can be a periodic time, such as 9:00-10:00 every day.
  • the model filter information supported by the first network element includes model filter information corresponding to the model generated by the first network element through federated learning.
  • the federated learning capability information of the first network element also includes at least one of the following:
  • the algorithm information based on the federated learning training of the first network element
  • Model accuracy information that can be achieved by federated learning training of the first network element
  • the model supported by the first network element can share information
  • the algorithm information based on the above-mentioned federated learning training of the first network element is the specific algorithm used for training the model, such as linear regression and neural network.
  • model accuracy information that can be achieved by the federated learning training of the first network element is used to indicate the accuracy of the model output result.
  • the speed information of the federated learning training of the first network element includes information on the time required for the model trained by the first network element based on federated learning to reach the target accuracy of the model.
  • model description method information supported by the first network element is used to indicate that the first network element supports a model description method corresponding to the model description method information based on the model description method.
  • model description method information may also be called model description method requirement information, or model description method expectation information.
  • the model description method information may be a model expression language represented by Open Neural Network Exchange (ONNX), etc., or a model framework represented by TensorFlow, Pytorch, etc.
  • ONNX Open Neural Network Exchange
  • TensorFlow TensorFlow
  • Pytorch a model framework represented by TensorFlow, Pytorch, etc.
  • sharable information of the model supported by the first network element is used to indicate that the first network element supports the shared model, where sharable means that the model can be interoperated, or sharable means that the model can be understood by each other. Shared means runnable.
  • model shareable information may also be called model shareable requirement information, or model shareable expected information.
  • model shareable information supported by the first network element includes one or more vendor information
  • other network elements belonging to the vendor information can interact with the first network element, such as obtaining the information trained by the first network element. Models and model information, etc.
  • model shareable information supported by the first network element includes one or more model information
  • other network elements can interact with the model trained by the first network element corresponding to the model information, for example, obtain the model information of the first network element. Trained models and model information, etc.
  • model shareable information can also be based on data analysis task granularity, that is, for each type of data analysis task, there is different shareable information (different manufacturer information, model information, etc.).
  • the above manufacturer information of the first network element is used to indicate the corresponding manufacturer of the first network element.
  • the above-mentioned manufacturer information of the first network element is used by the second network element to subsequently determine whether the model generated by the first network element based on federated learning training can be shared with other network elements.
  • the manufacturer information of other network elements is the same as the manufacturer information of the first network element, other network elements can share the model generated by the first network element based on federated learning training.
  • the manufacturer information of the first network element can be used by the second network element to select network elements that can participate in federated learning training.
  • the quantity information of the federated learning models supported by the first network element is used to indicate the quantity information of the federated learning training-based models supported by the first network element for each data analysis identifier, such as the maximum number and the minimum number. quantity.
  • the above-mentioned registration request also includes at least one of the following: type information of the first network element, first network element Supported service names.
  • the first network element is the NWDAF network element
  • the service name supported by the NWDAF network element is Nnwdaf_AnalyticsInfo_Request.
  • the type information of the first network element is used to indicate what kind of network element is registered this time.
  • the type information of the first network element is the NWDAF network element type, indicating the type of network element registered this time.
  • the network element is the NWDAF network element.
  • step 202 after the above-mentioned step 202, the following steps 203 and 204 are also included.
  • Step 203 The second network element sends a registration response to the first network element.
  • the above-mentioned registration response is used to indicate that the capability registration of the first device is successful.
  • the above-mentioned second network element can send an Nnrf_NFManagement_NFRegister response to the second network element to notify the first network element that the registration is successful.
  • Step 204 The first network element receives the registration response sent by the second network element.
  • Embodiments of this application provide a network element registration method.
  • the first network element can register the federated learning capability information of the first network element into the second network element through a registration request, so as to register the first network element into the second network element. in the network element, so that other network elements can find the first network element that can perform federated learning training through the second network element, so as to solve the problem of how the first network element registers the federated learning capability information to the second network element, and can be Issues found on other network elements.
  • the network element registration method provided by the embodiment of the present application further includes the following steps 205 and 206.
  • Step 205 The third network element sends a search request to the second network element.
  • the above search request is used to request to search for network elements that can perform federated learning training.
  • Step 206 The second network element receives the search request sent by the third network element.
  • the above search request includes first information
  • the first information includes at least one of the following: data analysis identification information corresponding to the target task, type information of federated learning training corresponding to the target task, federation corresponding to the target task
  • the time information of learning and training the metadata information of the network elements corresponding to the federated learning training corresponding to the target task, the network element type requirement information, and the network element quantity requirement information.
  • the network element type requirement information is used to indicate that the required search can be performed.
  • the network element quantity requirement information is used to indicate that the required network element can perform federated learning. The number of network elements required for training.
  • the above-mentioned third network element is a Model Training Logical Function (MTLF) network element, that is, a model training network element, which can be understood as a network element that initiates federated learning.
  • MTLF Model Training Logical Function
  • the above-mentioned third network element may send an Nnrf_NFDiscovery_Request search request to the second network element to request to find a network element that can perform federated learning training.
  • the third network element when the third network element does not support/does not perform independent training to generate model information corresponding to the target task, the third network element sends a search request to the second network element.
  • the above-mentioned third network element does not support/do not perform independent training to generate model information corresponding to the target task, which may be the case where the amount of data of the third network element is insufficient, or the third network element trains the model.
  • the accuracy rate is not up to standard.
  • the third network element may be a network element with federated learning server capabilities, and the third network element may find a network element with federated learning client capabilities to perform federated learning training.
  • the above-mentioned third network element may be a network element with federated learning client capabilities.
  • the third network element may search for a network element with federated learning server capabilities to request a network element with federated learning server capabilities.
  • Network elements are organized to conduct federated learning training.
  • the type information of the federated learning training corresponding to the above target task is used to instruct the acquisition of network element information of the network element that supports the type of federated learning training corresponding to the target task.
  • the third network element determines that federated learning training is required. , then instructs to obtain the network element information of the network element that supports federated learning training.
  • This network element information is used to instruct the network element to support federated learning training; or, if the third network element determines that horizontal federated learning training is needed, it instructs to obtain the network element that supports horizontal federation.
  • Network element information of the network element being learned and trained. This network element information is used to indicate that the network element supports horizontal federated learning training.
  • time information of federated learning training corresponding to the above target task is used to indicate the time information of the third network element planning to perform federated learning training, in order to find network elements that support federated learning training within this time range.
  • the information is a time period in the future, for example: 10 pm to 6 am the next morning.
  • the metadata information of the network elements trained by federated learning corresponding to the above target tasks is used to indicate the data information that the target network elements can cover, obtain, and provide.
  • the metadata information of the network elements trained by federated learning corresponding to the target task includes at least one of the following: input data type, output data type, data amount, and data range.
  • the above-mentioned first information also includes second information, which is used to request the acquisition of multiple required network elements capable of performing federated learning training.
  • the above network element quantity requirement information may include at least one of the following: minimum quantity requirement information, recommended quantity information, and maximum quantity requirement information.
  • the above-mentioned first information also includes at least one of the following:
  • Model accuracy information that can be achieved by federated learning training corresponding to the target task
  • Models supported by network elements trained by federated learning corresponding to the target task can share information
  • the algorithm information based on the federated learning training corresponding to the above target tasks is used to indicate the specific algorithm used in the training model, such as linear regression and neural network.
  • model accuracy information that can be achieved by federated learning training corresponding to the above target tasks is used to indicate the accuracy of the model output results.
  • the speed information of the federated learning training corresponding to the above target task is used to indicate the time required for the network element based on the federated learning trained model to reach the target accuracy of the model.
  • model description method information supported by the federated learning training network element corresponding to the above target task is used to indicate that the network element needs to support the model description method representation model based on the model description method information corresponding to the model description method information.
  • model shareable information supported by network elements trained by federated learning corresponding to the above target tasks is used to indicate that network elements need to support shared models.
  • the manufacturer information of the network element trained by federated learning corresponding to the above target task is used to indicate the corresponding manufacturer of the network element.
  • the information on the number of federated learning models supported by the federated learning trained network element corresponding to the above target task is used to indicate the number of federated learning trained models that the network element needs to support for each data analysis identifier.
  • the above-mentioned first information also includes at least one of the following: service information of network elements trained by federated learning corresponding to the target task, type information of network elements trained by federated learning corresponding to the target task, sense Area of interest information.
  • the service information of the federated learning trained network element corresponding to the above target task may be Nnwdaf_AnalyticsInfo_Request.
  • the type information of the network element of the federated learning training corresponding to the above target task can be the NWDAF network element type information.
  • the above-mentioned area of interest information may be in the form of TA(s), cell(s) or other expressions, used to indicate a request to discover network elements participating in federated learning training in the area of interest.
  • the network element registration method provided by the embodiment of the present application further includes the following step 207.
  • Step 207 The second network element determines the target network element according to the search request.
  • the federated learning capability information of the target network element matches the first information.
  • the second network element determines the network element whose federated learning capability information matches the first information as the target network element (one or more network elements).
  • step 207 can be specifically implemented through the following steps 207a to 207f.
  • Step 207a The second network element analyzes the identification information according to the data corresponding to the target task and determines the target network element.
  • the above-mentioned target network element supports federated learning training corresponding to the data analysis identification information.
  • the second network element determines the network element that supports federated learning training corresponding to the data analysis identification information as the target network element.
  • Step 207b The second network element determines the target network element according to the type of federated learning training corresponding to the target task.
  • the above-mentioned target network element supports the type of federated learning training corresponding to the target task.
  • the second network element determines the network element that supports the type of federated learning training corresponding to the target task as the target network element.
  • the second network element may determine a network element that supports federated learning training corresponding to the target task as the target network element.
  • the second network element may determine the network element that supports the federated learning server capability corresponding to the target task as the target network element.
  • the second network element may determine the network element that supports the federated learning client capability corresponding to the target task as the target network element.
  • Step 207c The second network element determines the target network element based on the time information of federated learning training corresponding to the target task.
  • the above-mentioned target network element supports federated learning training at a time corresponding to the target time information
  • the target time information is the time information of federated learning training corresponding to the target task.
  • the second network element determines the network element that supports federated learning training at the time corresponding to the target time information as the target network element.
  • Step 207d The second network element determines the target network element based on the metadata information of the federated learning trained network element corresponding to the target task.
  • the above-mentioned target network element supports the metadata information of the network element trained by federated learning corresponding to the target task.
  • the second network element determines the network element that supports federated learning training corresponding to the target task as the target network element.
  • Step 207e The second network element determines the target network element according to the network element quantity requirement information.
  • the number of the above target network elements meets the network element quantity requirement information.
  • the second network element finds the number of network elements that can perform federated learning training and meets the number of network elements indicated by the network element quantity requirement information
  • the network element that meets the quantity requirement can be determined as the target network element.
  • Step 207f The second network element determines the target network element according to the network element type requirement information.
  • the network element type of the above-mentioned target network element meets the network element type requirement information.
  • the second network element determines the network element that meets the network element type requirement information as the target network element.
  • the second network element determines the target task based on the algorithm information based on the federated learning training corresponding to the target task. Standard network element.
  • the second network element determines the network element that supports the algorithm based on the federated learning training corresponding to the target task as the target network element.
  • the second network element determines the target network element based on the model accuracy information that can be achieved by federated learning training corresponding to the target task.
  • the second network element determines the network element that can achieve the accuracy of the federated learning training model corresponding to the target task as the target network element.
  • the second network element determines the target network element based on the speed information of federated learning training corresponding to the target task.
  • the second network element determines the network element that supports the speed of federated learning training corresponding to the target task as the target network element;
  • the second network element determines the target network element based on the model description method information supported by the federated learning trained network element corresponding to the target task.
  • the second network element determines the network element that supports the model description method corresponding to the above model description method information (that is, the model description method information supported by the federated learning training network element corresponding to the target task) as the target network element.
  • the second network element determines the target network element based on the model shareable information supported by the federated learning trained network element corresponding to the target task.
  • the second network element determines the network element that supports the sharing model as the target network element.
  • the second network element determines the target network element based on the manufacturer information of the network element trained by federated learning corresponding to the target task.
  • the second network element determines the network element corresponding to the manufacturer information as the target network element.
  • the second network element determines the target network element based on information on the number of federated learning models supported by the federated learning trained network element corresponding to the target task.
  • the second network element determines the network element that supports the quantity information of the above federated learning model (that is, the quantity information of the federated learning model supported by the federated learning training network element corresponding to the target task) as the target network element.
  • the network element registration method provided by the embodiment of the present application further includes the following step A1.
  • Step A1 When the number of network elements found by the second network element that can perform federated learning training is less than the number of network elements indicated by the network element quantity requirement information, the second network element will find the network elements that can perform federated learning training. The network element is determined as the target network element.
  • the network elements found above that are capable of federated learning training are those that meet other required information of the third network element (for example, network element type, federated learning training time, etc.), but do not meet the required information on the number of network elements. Yuan.
  • the second network element when the number of network elements capable of federated learning training found by the second network element is equal to the number of network elements indicated by the network element quantity requirement information, the second network element will search The obtained network element that can perform federated learning training is determined as the target network element.
  • the network elements found above that are capable of federated learning training are network elements that meet the information requirements of the third network element.
  • the second network element when the number of network elements that the second network element finds that can perform federated learning training is greater than the number of network elements indicated by the network element quantity requirement information, the second network element will search Some of the obtained network elements that can perform federated learning training are determined as target network elements.
  • the second network element can sort all the found network elements that can perform federated learning training according to the preset internal logic, and then select the network elements that meet the required information on the number of network elements according to the sorting results. network element as the target network element.
  • the network element registration method provided by the embodiment of the present application further includes the following steps 208 and 209.
  • Step 208 The second network element sends a search response to the third network element.
  • the above search response includes identification information or address information of the target network element.
  • Step 209 The third network element receives the search response sent by the second network element.
  • the identification information of the target network element may be any of the following: FQDN information or IP address information.
  • the above-mentioned second network element may send an Nnrf_NFDiscovery_Request search response to the third network element to send network element information of network elements capable of federated learning training.
  • the above search response also includes second information
  • the second information includes at least one of the following: type information of federated learning training supported by the target network element, and time when the target network element supports federated learning training.
  • Information metadata information owned by the target network element.
  • the type information of federated learning training supported by the target network element is used to indicate the type of AI model training algorithm supported by the target network element.
  • the time information of the above-mentioned target network element supporting federated learning training is used to indicate the federated learning training time supported by the target network element.
  • Federated learning training performed within this time period can achieve better performance, for example: 10pm to 6am the next morning.
  • the metadata information possessed by the target network element is used to indicate the data information that the target network element can cover, obtain, and provide.
  • the metadata information of the target network element includes at least one of the following: input data type, output data type, data amount, and data range.
  • the above search response also includes type information of the target network element.
  • the type information of the target network element is used to indicate what kind of network element the target network element is.
  • the type information of the target network element is an NWDAF network element type, indicating that the target network element is an NWDAF network. Yuan.
  • the search response when the number of network elements capable of federated learning training found by the second network element is less than the number of network elements indicated by the network element quantity requirement information, the search response also includes the third Third information, the third information includes at least one of the following:
  • the first indication information is used to indicate that the number of target network elements is less than the number of network elements indicated by the network element quantity requirement information;
  • the fourth information is used to advise the third network element to delay searching for network elements that can perform federated learning training.
  • the fourth information includes: time information for delaying searching for network elements that can perform federated learning training.
  • the above time information may be a specific time.
  • the above fourth information may include time information "half an hour", so that when the third network element receives the third information, it can search for network elements capable of federated learning training again after half an hour based on the time information.
  • the second network element can use the found network elements that can perform federated learning training.
  • the network element being learned and trained is determined as the target network element, and the identification information or address information of the target network element is carried in the search response; on the other hand, the second network element can carry the third information in the search response to inform the third network At least one of the following elements: the number of target network elements does not meet the number of network elements required by the information indication, the number of target network elements, and it is recommended that the third network element can postpone the search for network elements that can perform federated learning training.
  • the search response may also include Fifth information, the fifth information is used to indicate that the number of target network elements meets the number of network elements indicated by the network element number requirement information.
  • the second network element can use the found network elements that can perform federated learning training.
  • the network element learned and trained is determined as the target network element, and the identification information or address information of the target network element is carried in the search response; on the other hand, the second network element can carry the fifth information in the search response to inform the third network
  • the number of target network elements meets the number indicated by the network element quantity requirement information.
  • the search response may also include Sixth information, the sixth information includes at least one of the following:
  • the third indication information is used to indicate that the number of found network elements capable of federated learning training is greater than the number of network elements indicated by the network element quantity requirement information;
  • the second network element can select from the network elements that can perform federated learning training. Select some of the network elements that meet the information required for the number of network elements as the target network elements, and carry the identification information or address information of the target network element in the search response; on the other hand, the second network element can carry the third network element in the search response.
  • the third network element after the third network element receives the third information, it can obtain the model information corresponding to the target task through federated learning training with the target network element; it may also not be determined with the second network element.
  • the target network element performs federated learning training, that is, it stops performing federated learning training; it can also delay and re-initiate a search request based on the fourth information or internal logic to re-search for network elements that can perform federated learning training and perform federated learning.
  • the third network element after the third network element receives the sixth information, it can select all or part of the network elements from the target network elements to perform federated learning training to obtain model information corresponding to the target task; it can also be based on The internal logic selects some network elements from all the network elements that can be federated learning and trained by the second network element to perform federated learning training to obtain model information corresponding to the target task.
  • the network element registration method provided by the embodiment of the present application further includes the following step 209a.
  • Step 209a The third network element selects all or part of the target network elements to perform federated learning training.
  • all or part of the above network elements can serve as a client for federated learning and training or a server for federated learning and training; specifically, the third network element can request information from the network element based on the network element type and network element type. Select the client for federated learning training or the server for federated learning training in the target network element.
  • the above-mentioned third network element may select all or Some network elements perform federated learning training.
  • the embodiment of the present application provides a network element registration method.
  • the second network element can receive a search request sent by the third network element to determine the target network element that can perform federated learning training according to the search request, and then can send the request to the third network element.
  • a search response is sent, so that the third network element determines a target network element that can perform federated learning training, and performs federated learning training with the target network element that can perform federated learning training to obtain model information corresponding to the target task.
  • the execution subject may also be a network element registration device, or a control module in the network element registration device for executing the network element registration method.
  • FIG. 9 shows a flow chart of a model determination method provided by this embodiment of the present application.
  • the model determination method provided by the embodiment of the present application may include the following steps 301 and 302.
  • Step 301 The third network element sends a search request to the second network element.
  • the above search request is used to request to search for network elements that can perform federated learning training.
  • Step 302 The second network element receives the search request sent by the third network element.
  • the third network element when the third network element does not support/does not perform independent training to generate model information corresponding to the target task, the third network element sends a search request to the second network element.
  • the above-mentioned third network element does not support/do not perform independent training to generate model information corresponding to the target task, which may be the case where the amount of data of the third network element is insufficient, or the third network element trains the model.
  • the accuracy rate is not up to standard.
  • the above-mentioned third network element is a Model Training Logical Function (MTLF) network element, that is, a model training network element, which can be understood as a network element that initiates federated learning.
  • MTLF Model Training Logical Function
  • the above-mentioned third network element may send an Nnrf_NFDiscovery_Request search request to the second network element to request to find a network element that can perform federated learning training.
  • the above search request includes first information
  • the first information includes at least one of the following: data analysis identification information corresponding to the target task, federated learning training type information corresponding to the target task, federated learning corresponding to the target task Training time information, metadata information of network elements trained by federated learning corresponding to the target task, network element type requirement information, and network element quantity requirement information.
  • the network element type requirement information is used to indicate that the required network element type requirement information is capable of federation.
  • the network element type includes the federated learning server network element type and/or the federated learning client network element type.
  • the network element quantity requirement information is used to indicate that the required network element can be used for federated learning training. The number of network elements required.
  • the type information of the federated learning training corresponding to the target task is used to indicate obtaining the network element information of the network element that supports the type of federated learning training corresponding to the target task, for example: the third network element determines If federated learning training is required, it is instructed to obtain the network element information of the network element that supports federated learning training.
  • the network element information is used to indicate that the network element supports federated learning training; or if the third network element determines that horizontal federated learning training is needed, then Instructs to obtain the network element information of the network element that supports horizontal federated learning training. This network element information is used to indicate that the network element supports horizontal federated learning training.
  • the time information of federated learning training corresponding to the above target task is used to indicate the time information of the third network element planning to perform federated learning training, so as to find the time information that supports federated learning training within this time range.
  • the time information is a time period in the future, for example: 10 pm to 6 am the next morning.
  • the metadata information of the federated learning-trained network element corresponding to the above target task is used to indicate the data information that the target network element can cover, obtain, and provide.
  • the metadata information of the network elements trained by federated learning corresponding to the target task includes at least one of the following: input data type, output data type, data amount, and data range.
  • the above-mentioned input data type refers to the input data type of federated learning training
  • the above-mentioned output data type refers to the output data type of federated learning training
  • the above data type refers to whether the input data and output data of federated learning training have obvious characteristics, such as gathering in the morning and spreading out in the evening.
  • the above data range refers to the service range of the target network element.
  • the above-mentioned data range includes at least one of the following: the service area of the target network element, the area where the target network element can collect data, and the objects from which the target network element can collect data.
  • the data collectable by the target network element includes metadata information and training data used for model training.
  • the area where the target network element can collect data can be any of the following: the target network element service area, the sub-area range under the target network element service area, and more fine-grained data can be collected.
  • Granular objects for example: a UE list).
  • the above-mentioned first information also includes at least one of the following:
  • Model accuracy information that can be achieved by federated learning training corresponding to the target task
  • Models supported by network elements trained by federated learning corresponding to the target task can share information
  • the algorithm information based on the federated learning training corresponding to the above target task is used to indicate the specific algorithm used in the training model, such as linear regression and neural network.
  • the model accuracy information that can be achieved by federated learning training corresponding to the above target task is used to indicate the accuracy level that the model output result needs to achieve.
  • the speed information of the federated learning training corresponding to the above target task is used to indicate the time required for the network element to reach the target accuracy of the model based on the federated learning training model.
  • the above-mentioned first information also includes at least one of the following: service information of network elements trained by federated learning corresponding to the target task, type information of network elements trained by federated learning corresponding to the target task, sense Area of interest information.
  • the service information of the federated learning trained network element corresponding to the above target task may be Nnwdaf_AnalyticsInfo_Request.
  • the type information of the network element of the federated learning training corresponding to the above target task can be the NWDAF network element type information.
  • the above-mentioned area of interest information may be in the form of TA(s), cell(s) or other expressions, used to indicate a request to discover network elements participating in federated learning training in the area of interest.
  • the embodiment of the present application provides a method for determining a model.
  • the third network element can search the second network element for network elements that can perform federated learning training through a search request. Since the search request includes the first information, the third network element can The target network element matching the first information is found. Thus, federated learning training is performed with the found target network elements that can be federated learning training.
  • the model determination method provided by the embodiment of the present application further includes the following steps 303 to 305 .
  • Step 303 The second network element sends a search response to the third network element.
  • the above search response includes identification information or address information of the target network element, and the target network element is a network element that supports federated learning training.
  • the identification information of the target network element may be any of the following: FQDN information or IP address information.
  • Step 304 The third network element receives the search response sent by the second network element.
  • the above-mentioned second network element may send an Nnrf_NFDiscovery_Request search response to the third network element to send network element information of network elements capable of federated learning training.
  • the above search response also includes second information
  • the second information includes at least one of the following: type information of federated learning training supported by the target network element, and time when the target network element supports federated learning training.
  • Information metadata information owned by the target network element.
  • the type information of federated learning training supported by the target network element is used to indicate the type of AI model training algorithm supported by the target network element.
  • the time information of the above-mentioned target network element supporting federated learning training is used to indicate the execution time of federated learning training supported by the target network element.
  • Federated learning training can be performed within this time period to achieve better results. Performance, for example: 10pm to 6am the next morning.
  • the metadata information possessed by the target network element is used to indicate the data information that the target network element can cover, obtain, and provide.
  • the metadata information of the target network element includes at least one of the following: input data type, output data type, data amount, and data range.
  • the above search response also includes type information of the target network element.
  • the type information of the target network element is used to indicate what kind of network element the target network element is.
  • the type information of the target network element is an NWDAF network element type, indicating that the target network element is an NWDAF network. Yuan.
  • Step 305 The third network element performs federated learning training with the target network element to obtain model information corresponding to the target task.
  • the third network element can determine and search for the target network element participating in the federated learning training based on the information included in the received search response, so as to perform federated learning with the found target network element.
  • the model determination method provided by the embodiment of the present application further includes the following steps 306 and 307 .
  • Step 306 The fourth network element sends a model request to the third network element.
  • the above model request is used to request model information corresponding to the target task.
  • Step 307 The third network element receives the model request sent by the fourth network element.
  • the fourth network element is a model inference network element, and the network element may be an AnLF network element, a NWDAF network element, or an NWDAF containing AnLF network element.
  • the fourth network element may send an Nnwdaf_MLMoldelInfo_Request model request to the third network element to request the third network element to feed back model information that meets the task.
  • the above model request includes at least one of the following: data analysis identification information corresponding to the target task, and limit information corresponding to the target task.
  • the data analysis identification information corresponding to the above target task may be user mobility trajectory (UE mobility) information.
  • UE mobility user mobility trajectory
  • the limitation information corresponding to the above target task is used to indicate the specific limitation conditions of the target task, such as: limiting the time, location and other information required for the target task.
  • the above model request also includes at least one of the following: indication information of the target network element and report limitation information.
  • the above indication information of the target network element is used to indicate the target network element of the target task.
  • identity information such as SUPI of the target network element can be specified.
  • the above report definition information is used to indicate the information, format, etc. that need to be reported for the target task, for example: the arrangement is in ascending order.
  • the model determination method provided by the embodiment of the present application further includes the following steps 308 and 309 .
  • Step 308 The third network element sends a model response to the fourth network element.
  • the above model response includes model information corresponding to the target task, and the model information is used by the fourth network element to perform model inference.
  • Step 309 The fourth network element receives the model response sent by the third network element.
  • the model information corresponding to the above target task includes at least one of the following: target information, federated learning training related information.
  • the above target information includes at least one of the following: model description information corresponding to the target task, model file corresponding to the target task, and model storage address information corresponding to the target task.
  • the model file corresponding to the target task includes at least one of the following: a complete network structure for generating model information corresponding to the target task, and parameter information for generating model information corresponding to the target task.
  • the above federated learning training related information includes at least one of the following: second instruction information, identification information of network elements participating in federated learning training, and capability information of network elements participating in federated learning training.
  • the above-mentioned second indication information is used to indicate that the model information corresponding to the target task is model information obtained through federated learning training.
  • the identification information of the network elements participating in federated learning training may be the identity and address information of the network elements, such as fully qualified domain name information, IP address information, etc.
  • the third network element may send the identification information of the network elements participating in the federated learning training to the fourth network element, so that the fourth network element can perform subsequent inference on the federated learning training.
  • the model information corresponding to the target task also includes at least one of the following: data analysis identification information corresponding to the target task, and definition information corresponding to the target task.
  • the limited information corresponding to the target task includes information about the area of interest or information about the target terminal.
  • the model determination method provided by the embodiment of the present application may also include the following step 401.
  • Step 401 When the third network element cannot obtain the training data of the target task corresponding to the area of interest or the target terminal, the third network element performs federated learning training for the target task.
  • the above-mentioned areas of interest are areas where the model trained based on federated learning needs/to be predicted information.
  • the above-mentioned target terminal is a terminal where the model based on federated learning training requires/to-be-predicted information.
  • the limited information corresponding to the target task includes information about the area of interest or information about the target terminal.
  • the model determination method provided by the embodiment of the present application may also include the following step 402.
  • Step 402 The third network element determines the metadata information of the federated learning trained network element corresponding to the target task based on the information of the area of interest or the information of the target terminal.
  • the above metadata information includes the data range of network elements trained by federated learning.
  • Embodiments of the present application provide a model determination method.
  • the third network element can receive a model request sent by the fourth network element and request model information corresponding to the target task. Then the third network element can search the second network element through a search request. For network elements capable of federated learning training, since the search request includes the first information, the third network element can find the target network element matching the first information. Through federated learning training with the found target network elements, the model information corresponding to the target task is obtained, thereby improving the success rate of model transfer.
  • the execution subject may also be a model determination device, or a control module in the model determination device for executing the model determination method.
  • the method provided by the embodiment of the present application includes the following steps 21 to 32.
  • Step 21 Intelligent devices in different domains (for example, NWDAF network elements on the core network side, base stations, UEs, third-party application servers, etc.) send capability storage devices such as NRF (for example, NRF network elements, UDM network elements, DCAF network Yuan, etc.) sends a capability registration message to perform capability registration.
  • NRF for example, NRF network elements, UDM network elements, DCAF network Yuan, etc.
  • the NWDAF network element can be registered through (Nnrf_NFManagement_NFRegister Register).
  • Step 22 The NRF network element stores the information of the NWDAF network element.
  • Step 23 The NRF network element sends a registration response message.
  • the NRF network element can respond with a (Nnrf_NFManagement_NFRegister response) message to notify the NWDAF network element of successful registration.
  • step 21 when the NWDAF network element sends the capability registration message to the NRF network element, in addition to its own identification information, supported analytic ID and other information, it also sends "supported training type information”, “supported federated learning time” and “metadata” Information Meta data” etc.
  • the information required for registration is:
  • NF instance ID, FQDN or IP address of NF network element instance identification information; refers to the network element identification information registered this time, for example, its FQDN information (Fully Qualified Domain Name, fully qualified domain name, used to indicate this network element location and connection to this network element) or IP address information (another type of identification information).
  • FQDN information Full Qualified Domain Name, fully qualified domain name, used to indicate this network element location and connection to this network element
  • IP address information another type of identification information
  • Names of supported NF services (if applicable), the service names supported by the network element; such as NWDAF network element, there are service names such as Nnwdaf_AnalyticsInfo_Request, etc.
  • the network element registration information must also contain at least one of the following:
  • Supported training type information refers to the type of AI model training algorithm supported by the network element, such as "federated learning”, “deep learning” "Xi” and other such information.
  • Supported federated learning time refers to the federated learning time supported by the network element. Federated learning can achieve better performance during this time period, such as "10 pm to 6 am the next morning”.
  • Metadata information refers to the data information that the network element can cover, obtain, and provide, including data type, data characteristics, data volume, and other information.
  • Analytics ID data analysis ID; such as "UE mobility” (user mobility trajectory), etc.
  • Algorithm used for model training specific algorithm used for training model, such as linear regression, neural network, etc.
  • Model training can achieve model accuracy; the accuracy of model output results.
  • the speed of model training used to indicate the time required for model training to reach a specific accuracy.
  • Step 24 The task consumer (consumer) sends a data analysis request message to the model inference network element AnLF (or NWDAF, NWDAF containing AnLF).
  • AnLF or NWDAF, NWDAF containing AnLF.
  • the task consumer (consumer) can send a data analysis request message through (Nnwdaf_AnalyticsInfo_Request or Nnwdaf_AnalyticsSubscription_Subscribe).
  • Step 25 After receiving the task request, AnLF sends a model request to MTLF. Request MTLF feedback on models that fit the mission.
  • AnLF can send a model request through (Nnwdaf_MLMoldelInfo_Request).
  • the request includes task description information, such as:
  • analytic ID data analysis ID; such as "UE mobility” (user mobility trajectory), etc.
  • Filter info indicates the specific qualification conditions of this task, such as the time, location and other information required to limit the task.
  • Target network element/UE and other information indicating the target network element or UE of the task, such as UE mobility, the SUPI and other identity information of the target UE can be specified.
  • Reporting info reporting limited information; specifying the information and format required for reporting by the task, such as ascending order.
  • Step 26 MTLF determines whether it can be independently trained to generate a model that meets the requirements based on the task description information in step 25. If it cannot be independently trained and generated, federated learning can be initiated.
  • the failure to independently train and generate may be due to insufficient number of data sets, insufficient distinction of data features, etc., so that the accuracy cannot reach the target accuracy, etc.
  • Step 27 The model training network element MTLF and other network element devices that initiate federated training send a request to the storage capability network element such as NRF to find network elements and other devices that can perform federated learning training.
  • the storage capability network element such as NRF
  • the model training network element MTLF can send a search request through (Nnrf_NFDiscovery_Request).
  • MTLF sends a request to NRF to find network elements participating in federated learning, which needs to include:
  • NF type of the target NF the type of the target network element; if the network element of the target service is NWDAF, the information can be NWDAF type.
  • Area of interest it can be TA(s), cell(s) or other expressions, used to indicate the request to discover network elements participating in federated learning in the area of interest.
  • MTLF When MTLF sends a request to NRF, it must also include "training type”, “training condition qualification information”, etc.
  • Training type refers to the training type supported by the network element device that is expected to be obtained.
  • the network element wants to perform federated learning, and it is hoped that the obtained network element information supports federated learning training, then the information can be Information indicating that the network element supports federated learning, such as "federated learning", etc.
  • Training condition limitation information refers to the screening information of network element equipment, etc., and further narrowing the scope of network element equipment, etc. to select network elements that are more suitable for needs, such as "federated learning time information", “required data information” "wait. Include at least one of the following:
  • Supports federated learning time refers to the target time information planned for federated learning, and the target is to select network element equipment that can perform federated learning within this time range.
  • the time information will be a time period in the future, such as "10 pm to 6 am the next morning”.
  • Required data information refers to the data information that is expected to be covered, obtained, and provided by the target network element, including data type, data characteristics, data volume, and other information.
  • Step 28 NRF returns network element and other device information that matches the information retrieved in step 27 to MTLF.
  • the capability storage network element such as NRF returns to the federated learning initiator the network element device information that meets the training condition definition information in step 26, and can return a response message through (Nnrf_NFDiscovery_Request).
  • Network element device identification information indicating the FQDN and IP address of the target network element and device.
  • Type information of network element equipment such as NWDAF.
  • Capability information of network element equipment including at least one of the following:
  • Training type refers to the training type supported by the network element device, such as "federated learning", etc.
  • Supported federated learning time refers to the federated learning time supported by the network element. Federated learning can achieve better performance during this time period, such as "10 pm to 6 am the next morning".
  • Metadata information refers to the data information that the network element can cover, obtain, and provide, including data type, data characteristics, data volume, and other information.
  • Step 29 MTLF determines and searches for network elements and other equipment participating in the training based on the information returned in step 28. MTLF performs federated learning with the found network elements and other equipment.
  • Step 30 After the MTLF training is completed, send the feedback model to AnLF, including task description information.
  • Task limitation information information that further describes task requirements, such as limited time, location and other information.
  • identification information and capability information about federated learning members including at least one of the following:
  • Identification information of members participating in federated learning can be the member’s identity and address information, such as FQDN or IP address, etc.
  • MTLF sends device information such as network elements participating in federated learning to AnLF for AnLF to perform subsequent federated learning inferences.
  • Capability information of members participating in federated learning including time information supporting federated learning and data information of the device itself, etc.
  • Step 31 After receiving the feedback model and other information, AnLF performs model inference.
  • AnLF can perform cooperative inference based on the obtained federated learning information, such as network elements participating in federated learning and other device information.
  • Step 32 AnLF feeds back the task report to the consumer based on the inference results.
  • Figure 14 shows a possible structural diagram of the network element registration device involved in the embodiment of the present application.
  • the network element registration device 80 may include: a sending module 81.
  • the sending module 81 is used to send a registration request to the second network element.
  • the registration request is used to request to register the federated learning capability information of the first network element to the second network element.
  • the federated learning capability information of the first network element includes At least one of the following: type information of federated learning training supported by the first network element, time information of federated learning training supported by the first network element, and metadata information possessed by the first network element.
  • Embodiments of the present application provide a network element registration device.
  • the network element registration device can register the federated learning capability information of the first network element into the second network element through a registration request, so as to register the first network element into the second network element. in the network element, so that other network elements can find the first network element that can perform federated learning training through the second network element, so as to solve the problem of how the first network element registers the federated learning capability information to the second network element, and can be Issues found on other network elements.
  • the type information of federated learning training supported by the first network element includes at least one of the following: indication information of whether the first network element supports federated learning training, horizontal federated learning training type, vertical federated learning Training type, federated learning server capabilities, federated learning client capabilities.
  • the metadata information of the first network element includes at least one of the following: input data type, output data type, data amount, and data range.
  • the above data range includes at least one of the following: a service area of the first network element, an area where the first network element can collect data, and an object from which the first network element can collect data.
  • the above-mentioned registration request also includes at least one of the following: identification information of the first network element, data analysis identification information supported by the first network element, model filter information supported by the first network element, Information about the network element type of the first network element.
  • the network element type includes the federated learning server network element type and/or the federated learning client network element type.
  • the above data analysis identification information corresponds to the type information of federated learning training.
  • the above data analysis identification information corresponds to the federated learning capability information of the first network element.
  • the above model filter information includes model filter information corresponding to the model generated by the first network element through federated learning.
  • the federated learning capability information of the first network element also includes at least one of the following: algorithm information on which the federated learning training of the first network element is based; the federated learning training of the first network element can reach model accuracy information; federated learning training speed information of the first network element; model description method information supported by the first network element; model shareable information supported by the first network element; manufacturer information of the first network element; Information about the number of federated learning models supported by the network element.
  • the speed information of the federated learning training of the first network element includes information on the time required for the model trained by the first network element based on federated learning to reach the target accuracy of the model.
  • the network element registration device provided by the embodiment of the present application can implement each process implemented by the first network element in the above method embodiment, and achieve the same technical effect. To avoid duplication, the details will not be described here.
  • Figure 15 shows a possible structural diagram of the network element registration device involved in the embodiment of the present application.
  • the network element registration device 90 may include: a receiving module 91.
  • the receiving module 91 is used to receive a registration request sent by the first network element.
  • the registration request is used to request to register the federated learning capability information of the first network element to the second network element.
  • the federated learning capability information of the first network element Including at least one of the following: type information of federated learning training supported by the first network element, time information of federated learning training supported by the first network element, and metadata information possessed by the first network element.
  • Embodiments of the present application provide a network element registration device.
  • the network element registration device can register the federated learning capability information of the first network element into the second network element by receiving a registration request sent by the first network element, so as to realize the registration of the first network element.
  • One network element is registered to the second network element, so that other network elements can find the first network element that can perform federated learning training, so as to solve the problem of how the first network element registers the federated learning capability information to the second network element, and can Problems found by other network elements.
  • the federated learning capability information of the first network element also includes at least one of the following:
  • the above-mentioned receiving module 91 is also used to receive a search request sent by a third network element.
  • the search request is used to request to find a network element that can perform federated learning training.
  • the search request includes the first information.
  • the first information includes at least one of the following: data analysis identification information corresponding to the target task, type information of federated learning training corresponding to the target task, time information of federated learning training corresponding to the target task, network of federated learning training corresponding to the target task.
  • the network element type requirement information is used to indicate the network element type corresponding to the network element that needs to be found and can perform federated learning training.
  • the network element The type includes the federated learning server network element type and/or the federated learning client network element type.
  • the network element quantity requirement information is used to indicate the quantity requirement of the required network elements that can perform federated learning training.
  • the above-mentioned first information also includes at least one of the following: algorithm information on which federated learning training corresponding to the target task is based; model accuracy information that can be achieved by federated learning training corresponding to the target task; target task Corresponding federated learning training speed information; model description method information supported by federated learning training network elements corresponding to the target task; model shareable information supported by federated learning training network elements corresponding to the target task; federated learning training corresponding to the target task Manufacturer information of the network element; information on the number of federated learning models supported by the federated learning training network element corresponding to the target task.
  • the network element registration device 90 provided by the embodiment of the present application further includes: a determination module.
  • the determining module is configured to determine the target network element according to the search request after the receiving module 91 receives the search request sent by the third network element, and the federated learning capability information of the target network element matches the first information.
  • the above determination module is specifically used to determine the target network element according to the data analysis identification information corresponding to the target task, wherein the target network element supports federated learning training corresponding to the data analysis identification information; according to the target task
  • the corresponding federated learning training type determines the target network element, where the target network element supports the federated learning training type corresponding to the target task; determines the target network element based on the time information of the federated learning training corresponding to the target task, where the target network element The element supports federated learning training at a time corresponding to the target time information.
  • the target time information is the time information of the federated learning training corresponding to the target task; the target is determined based on the metadata information of the federated learning training network element corresponding to the target task.
  • Network elements where the target network element supports the metadata information of the network elements trained by federated learning corresponding to the target task; determine the target network element based on the information required for the number of network elements, where the number of target network elements meets the requirement for the number of network elements Information; determine the target network element according to the network element type requirement information, where the network element type of the target network element meets the network element type requirement information.
  • the above-mentioned determination module is used to find the network capable of federated learning training on the second network element. If the number of network elements is less than the number of network elements indicated by the network element quantity requirement information, the found network element that can perform federated learning training is determined as the target network element.
  • the network element registration device 90 provided in the embodiment of the present application further includes: a sending module.
  • the sending module is configured to send a search response to the third network element after the determination module determines the target network element according to the search request.
  • the search response includes the identification information or address information of the target network element.
  • the above search response also includes second information
  • the second information includes at least one of the following: type information of federated learning training supported by the target network element, and time when the target network element supports federated learning training.
  • Information metadata information owned by the target network element.
  • the search response when the number of network elements capable of federated learning training found by the second network element is less than the number of network elements indicated by the network element quantity requirement information, the search response also includes a third Information, the third information includes at least one of the following: first indication information, the first indication information is used to indicate that the number of target network elements is less than the number of network elements indicated by the network element quantity requirement information; quantity information of the target network elements; The fourth information is used to advise the third network element to delay searching for network elements that can perform federated learning training.
  • the fourth information includes: time information for delaying searching for network elements that can perform federated learning training.
  • the network element registration device provided by the embodiment of the present application can implement each process implemented by the second network element in the above method embodiment, and achieve the same technical effect. To avoid duplication, the details will not be described here.
  • Figure 16 shows a possible structural schematic diagram of the model determination device involved in the embodiment of the present application.
  • the model determination device 100 may include: a sending module 101.
  • the sending module 101 is used to send a search request to the second network element.
  • the search request is used to request to find a network element that can perform federated learning training.
  • the search request includes first information, and the first information includes at least one of the following: : Data analysis identification information corresponding to the target task, type information of federated learning training corresponding to the target task, time information of federated learning training corresponding to the target task, metadata information of the network element corresponding to the federated learning training corresponding to the target task, network Element type requirement information and network element quantity requirement information.
  • the network element type requirement information is used to indicate the network element type corresponding to the network element that needs to be found and can perform federated learning training.
  • the network element type includes the federated learning server network element type and /or federated learning client network element type, network element quantity requirement information is used to indicate the required number of network elements capable of federated learning training.
  • Embodiments of the present application provide a model determination device.
  • the model determination device can search the second network element for network elements capable of federated learning training through a search request. Since the search request includes the first information, the third network element can search to the target network element matching the first information. Thus, federated learning training is performed with the found target network elements that can be federated learning training.
  • the metadata information of the network elements trained by federated learning corresponding to the above target task includes at least one of the following: input data type, output data type, data amount, and data range.
  • the above-mentioned first information also includes at least one of the following: algorithm information on which federated learning training corresponding to the target task is based; model accuracy information that can be achieved by federated learning training corresponding to the target task; target task The speed information of the corresponding federated learning training; the model description method information supported by the network element of the federated learning training corresponding to the target task; the federated learning corresponding to the target task
  • the models supported by the trained network elements can share information; the manufacturer information of the network elements trained by federated learning corresponding to the target task; and the number of federated learning models supported by the network elements trained by federated learning corresponding to the target task.
  • the model determination device 100 provided by the embodiment of the present application further includes: a receiving module and an acquisition module.
  • the receiving module is configured to receive a search response sent by the second network element after the sending module 101 sends a search request to the second network element.
  • the search response includes the identification information or address information of the target network element, and the target network element supports The network element trained by federated learning; the acquisition module is used to obtain model information corresponding to the target task through federated learning training with the target network element.
  • the above search response also includes second information
  • the second information includes at least one of the following: type information of federated learning training supported by the target network element, and time when the target network element supports federated learning training.
  • Information metadata information owned by the target network element.
  • the search response when the number of network elements capable of federated learning training found by the second network element is less than the number of network elements indicated by the network element quantity requirement information, the search response also includes a third Information, the third information includes at least one of the following: first indication information, the first indication information is used to indicate that the number of target network elements is less than the number of network elements indicated by the network element quantity requirement information; quantity information of the target network elements; The fourth information is used to advise the third network element to delay searching for network elements that can perform federated learning training.
  • the fourth information includes: time information for delaying searching for network elements that can perform federated learning training.
  • the model determination device 100 provided by the application embodiment further includes: a selection module; a selection module configured to select all target network elements after the receiving module receives the search response sent by the second network element. Or some network elements perform federated learning training.
  • the above-mentioned receiving module is also configured to receive a model request sent by the fourth network element before the sending module 101 sends a search request to the second network element.
  • the model request is used to request a model corresponding to the target task.
  • Information, the model request includes at least one of the following: data analysis identification information corresponding to the target task, and qualification information corresponding to the target task.
  • the above-mentioned sending module 101 is also used to send a model response to the fourth network element after the acquisition module obtains the model information corresponding to the target task through federated learning training with the target network element.
  • the response includes model information corresponding to the target task, and the model information is used by the fourth network element for model inference.
  • the model determination device 100 provided by the embodiment of the present application further includes: an execution module.
  • the limited information corresponding to the target task includes the information of the area of interest or the information of the target terminal; the execution module is used to execute the target task when the third network element cannot obtain the training data of the target task corresponding to the area of interest or the target terminal. Federated learning training for tasks.
  • the model determination device 100 provided by the embodiment of the present application further includes: a determination module.
  • the limited information corresponding to the target task includes the information of the area of interest or the information of the target terminal; the determination module is used to determine the elements of the network element of the federated learning training corresponding to the target task based on the information of the area of interest or the information of the target terminal.
  • Data information, this metadata information includes the data range of network elements trained by federated learning.
  • the model information corresponding to the target task includes at least one of the following: target information, federated learning training related information, and the target information includes at least one of the following: model description information corresponding to the target task, target task correspondence The model file and the model storage address information corresponding to the target task.
  • the above federated learning training related information includes at least one of the following: second indication information, identification information of network elements participating in federated learning training, capability information of network elements participating in federated learning training, second The indication information is used to indicate that the model information corresponding to the target task is model information obtained through federated learning training.
  • the model determination device provided by the embodiment of the present application can realize each process implemented by the third network element in the above method embodiment, and achieve the same technical effect. To avoid duplication, the details will not be described here.
  • this embodiment of the present application also provides a communication device 5000, including a processor 5001 and a memory 5002.
  • the memory 5002 stores programs or instructions that can be run on the processor 5001, for example , when the communication device 5000 is a network element, when the program or instruction is executed by the processor 5001, each step of the above-mentioned first network element side method embodiment is implemented, and the same technical effect can be achieved, or the above-mentioned second network element side can be achieved.
  • Each step of the method embodiment can achieve the same technical effect, or each step of the above third network element side method embodiment can be implemented and the same technical effect can be achieved. To avoid duplication, the details will not be described here.
  • An embodiment of the present application also provides a network element, including a processor and a communication interface.
  • the communication interface is used to send a registration request to the second network element.
  • the registration request is used to request to register the federated learning capability information of the first network element to the second network element.
  • the federated learning capability information of the first network element includes at least one of the following: the type information of federated learning training supported by the first network element, the time information of the federated learning training supported by the first network element, and the information of the federated learning training supported by the first network element. metadata information.
  • This network element embodiment corresponds to the above-mentioned first network element side method embodiment. Each implementation process and implementation manner of the above-mentioned method embodiment can be applied to this network element embodiment, and can achieve the same technical effect.
  • An embodiment of the present application also provides a network element, including a processor and a communication interface.
  • the communication interface is used to receive a registration request sent by the first network element.
  • the registration request is used to request to register the federated learning capability information of the first network element to
  • the federated learning capability information of the second network element and the first network element includes at least one of the following: type information of federated learning training supported by the first network element, time information of federated learning training supported by the first network element, information on the federated learning training supported by the first network element, Has metadata information.
  • This network element embodiment corresponds to the above-mentioned second network element side method embodiment.
  • Each implementation process and implementation method of the above-mentioned method embodiment can be applied to this network element embodiment, and can achieve the same technical effect.
  • Embodiments of the present application also provide a network element, including a processor and a communication interface.
  • the communication interface is used to send a search request to the second network element.
  • the search request is used to request to find a network element that can perform federated learning training.
  • the search request is in progress.
  • Including first information the first information includes at least one of the following: data analysis identification information corresponding to the target task, type information of federated learning training corresponding to the target task, time information of federated learning training corresponding to the target task, time information corresponding to the target task.
  • the metadata information, network element type requirement information, and network element quantity requirement information of the network elements trained by federated learning are examples of the network elements trained by federated learning.
  • the network element type requirement information is used to indicate the network element corresponding to the network element that needs to be found and can be federated learning training.
  • Type the network element type includes the federated learning server network element type and/or the federated learning client network element type, and the number of network elements
  • the requirement information is used to indicate the required number of network elements to be found that can perform federated learning training.
  • This network element embodiment corresponds to the above-mentioned third network element side method embodiment. Each implementation process and implementation method of the above-mentioned method embodiment can be applied to this network element embodiment, and can achieve the same technical effect.
  • FIG. 18 is a schematic diagram of the hardware structure of a network element that implements an embodiment of the present application.
  • the network element is a first network element, or the network element is a second network element, or the network element is a third network element.
  • the network element 1200 includes: a processor 1201, a network interface 1202, and a memory 1203.
  • the network interface 1202 is, for example, a universal public wireless interface.
  • the network element 1200 in the embodiment of the present application also includes: instructions or programs stored in the memory 1203 and executable on the processor 1201.
  • the processor 1201 calls the instructions or programs in the memory 1203 to execute the methods executed by each of the above modules. , and achieve the same technical effect. To avoid repetition, we will not repeat them here.
  • the network element provided by the embodiment of the present application can implement each process implemented by the first network element, the second network element and the third network element in the above method embodiment, and achieve the same technical effect. To avoid duplication, the details will not be described here.
  • Embodiments of the present application also provide a readable storage medium.
  • Programs or instructions are stored on the readable storage medium.
  • the program or instructions are executed by a processor, each process of the above method embodiments is implemented and the same technology can be achieved. The effect will not be described here to avoid repetition.
  • the processor is the processor in the communication device described in the above embodiment.
  • the readable storage medium includes computer readable storage media, such as computer read-only memory ROM, random access memory RAM, magnetic disk or optical disk, etc.
  • An embodiment of the present application further provides a chip.
  • the chip includes a processor and a communication interface.
  • the communication interface is coupled to the processor.
  • the processor is used to run programs or instructions to implement various processes of the above method embodiments. , and can achieve the same technical effect, so to avoid repetition, they will not be described again here.
  • chips mentioned in the embodiments of this application may also be called system-on-chip, system-on-a-chip, system-on-chip or system-on-chip, etc.
  • Embodiments of the present application further provide a computer program/program product.
  • the computer program/program product is stored in a storage medium.
  • the computer program/program product is executed by at least one processor to implement each of the above method embodiments.
  • the process can achieve the same technical effect. To avoid repetition, it will not be described again here.
  • An embodiment of the present application also provides a communication system, including: a first network element, a second network element and a third network element.
  • the first network element can be used to perform the steps of the network element registration method as described above, so
  • the second network element may be used to perform the steps of the network element registration method as described above, and the third network element may be used to perform the steps of the model determination method as described above.
  • the methods of the above embodiments can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is better. implementation.
  • the technical solution of the present application can be embodied in the form of a computer software product that is essentially or contributes to the existing technology.
  • the computer software product is stored in a storage medium (such as ROM/RAM, disk , CD), including several instructions to cause a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of this application.

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本申请公开了一种网元注册方法、模型确定方法、装置、网元、通信系统及存储介质，本申请实施例的网元注册方法包括：第一网元向第二网元发送注册请求，该注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元，第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。

Description

网元注册方法、模型确定方法、装置、网元、通信系统及存储介质
相关申请的交叉引用
本申请主张在2022年03月28日在中国提交的中国专利申请号202210317250.2的优先权，以及在2022年11月22日在中国提交的中国专利申请号202211468901.4的优先权，其全部内容通过引用包含于此。
技术领域
本申请属于通信技术领域，具体涉及一种网元注册方法、模型确定方法、装置、网元、通信系统及存储介质。
背景技术
随着通信技术的不断发展，出现了联邦学习技术，联邦学习可以在一个中心服务器的协调下让多个客户端互相合作，以得到一个完整的机器学习模型。然而，现有技术中，没有对联邦学习这种人工智能算法进行额外的增强或者区别对待，例如：默认能够进行训练的网元、基站、用户设备(User Equipment，UE)等参与者都能参与联邦学习，如此这种未进行筛选的联邦学习的对象选择，由于各种原因(例如UE网络信号较差等)，可能会造成联邦学习对象丢失，学习效率降低等问题。因此，如何增强联邦学习中网元注册流程是亟待解决的问题。
发明内容
本申请实施例提供一种网元注册方法、模型确定方法、装置、网元、通信系统及存储介质，能够解决如何增强联邦学习中网元注册流程的问题。
第一方面，提供了一种网元注册方法，该方法包括：第一网元向第二网元发送注册请求，该注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元，第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。
第二方面，提供了一种网元注册装置，应用于第一网元，该网元注册装置包括：发送模块。发送模块，用于向第二网元发送注册请求，该注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元，第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。
第三方面，提供了一种网元注册方法，该方法包括：第二网元接收第一网元发送的注册请求，该注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元，第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。
第四方面，提供了一种网元注册装置，应用于第二网元，该网元注册装置包括：接收模块。接收模块，用于接收第一网元发送的注册请求，该注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元，第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。
第五方面，提供了一种模型确定方法，该方法包括：第三网元向第二网元发送查找请求，该查找请求用于请求查找能够进行联邦学习训练的网元，查找请求中包括第一信息，该第一信息包括以下至少一项：目标任务对应的数据分析标识信息、目标任务对应的联邦学习训练的类型信息、目标任务对应的联邦学习训练的时间信息、目标任务对应的联邦学习训练的网元所具有的元数据信息、网元类型要求信息、网元数量要求信息，该网元类型要求信息用于指示所需查找的能够进行联邦学习训练的网元对应的网元类型，该网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型，网元数量要求信息用于指示所需查找的能够进行联邦学习训练的网元的数量要求。
第六方面，提供了一种模型确定装置，该模型确定装置包括：发送模块。发送模块，用于向第二网元发送查找请求，该查找请求用于请求查找能够进行联邦学习训练的网元，查找请求中包括第一信息，该第一信息包括以下至少一项：目标任务对应的数据分析标识信息、目标任务对应的联邦学习训练的类型信息、目标任务对应的联邦学习训练的时间信息、目标任务对应的联邦学习训练的网元所具有的元数据信息、网元类型要求信息、网元数量要求信息，该网元类型要求信息用于指示所需查找的能够进行联邦学习训练的网元对应的网元类型，该网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型，网元数量要求信息用于指示所需查找的能够进行 联邦学习训练的网元的数量要求。
第七方面，提供了一种网元，该网元包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如第一方面所述的网元注册方法的步骤。
第八方面，提供了一种网元，包括处理器及通信接口，其中，所述通信接口用于向第二网元发送注册请求，该注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元，第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。
第九方面，提供了一种网元，该网元包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如第三方面所述的网元注册方法的步骤。
第十方面，提供了一种网元，包括处理器及通信接口，其中，所述通信接口用于接收第一网元发送的注册请求，该注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元，第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。
第十一方面，提供了一种网元，该网元包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如第五方面所述的模型确定方法的步骤。
第十二方面，提供了一种网元，包括处理器及通信接口，其中，所述通信接口用于向向第二网元发送查找请求，该查找请求用于请求查找能够进行联邦学习训练的网元，查找请求中包括第一信息，该第一信息包括以下至少一项：目标任务对应的数据分析标识信息、目标任务对应的联邦学习训练的类型信息、目标任务对应的联邦学习训练的时间信息、目标任务对应的联邦学习训练的网元所具有的元数据信息、网元类型要求信息、网元数量要求信息，该网元类型要求信息用于指示所需查找的能够进行联邦学习训练的网元对应的网元类型，该网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型，网元数量要求信息用于指示所需查找的能够进行联邦学习训练的网元的数量要求。
第十三方面，提供了一种通信系统，包括：如第二方面所述的网元注册装置、如第四方面所述的网元注册装置和第六方面所述的模型确定装置；或者，包括：如第七方面、第九方面和第十一方面所述的网元；或者，包括：如第八方面、第十方面和第十二方面所述的网元。其中，所述网元可用于执行如第一方面、第三方面和第五方面所述的方法的步骤。
第十四方面，提供了一种可读存储介质，所述可读存储介质上存储程序或指令，所述程序或指令被处理器执行时实现如第一方面所述的方法的步骤，或者实现如第三方面所述的方法的步骤，或者实现如第五方面所述的方法的步骤。
第十五方面，提供了一种芯片，所述芯片包括处理器和通信接口，所述通信接口和所述处理器耦合，所述处理器用于运行程序或指令，实现如第一方面所述的方法，或实现如第三方面所述的方法，或实现如第五方面所述的方法。
第十六方面，提供了一种计算机程序/程序产品，所述计算机程序/程序产品被存储在存储介质中，所述计算机程序/程序产品被至少一个处理器执行以实现如第一方面所述的网元注册方法的步骤，或者实现如第三方面所述的网元注册方法的步骤，或者实现如第五方面所述的模型确定方法的步骤。
在本申请实施例中，第一网元可以通过注册请求，将第一网元的联邦学习能力信息注册到第二网元中，以实现将第一网元注册到第二网元中，从而使得其他网元能够通过第二网元查找到能够进行联邦学习训练的第一网元，以解决第一网元如何将联邦学习能力信息注册到第二网元，并能够被其他网元查找到的问题。
第二网元可以通过接收第一网元发送的注册请求，将第一网元的联邦学习能力信息注册到第二网元中，以实现将第一网元注册到第二网元中，从而使得其他网元查找到能够进行联邦学习训练的第一网元，以解决第一网元如何将联邦学习能力信息注册到第二网元，并能够被其他网元查找到的问题。
第三网元可以通过查找请求，向第二网元查找能够进行联邦学习训练的网元，由于查找请求中包括第一信息，因此第三网元可以查找到与第一信息匹配的目标网元。从而与查找到的能够进行联邦学习训练的目标网元进行联邦学习训练。
附图说明
图1是本申请实施例提供的一种无线通信系统的架构示意图；
图2是本申请实施例提供的一种横向联邦学习的原理示意图；
图3是本申请实施例提供的一种神经网络的示意图；
图4是本申请实施例提供的一种神经元的示意图；
图5是本申请实施例提供的一种网元注册方法的流程图之一；
图6是本申请实施例提供的一种网元注册方法的流程图之二；
图7是本申请实施例提供的一种网元注册方法的流程图之三；
图8是本申请实施例提供的一种网元注册方法的流程图之四；
图9是本申请实施例提供的一种模型确定方法的流程图之一；
图10是本申请实施例提供的一种模型确定方法的流程图之二；
图11是本申请实施例提供的一种模型确定方法的流程图之三；
图12是本申请实施例提供的一种模型确定方法的流程图之四；
图13是本申请实施例提供的一种网元注册方法及模型确定方法的流程图；
图14是本申请实施例提供的一种网元注册装置的结构示意图之一；
图15是本申请实施例提供的一种网元注册装置的结构示意图之二；
图16是本申请实施例提供的一种模型确定装置的结构示意图；
图17是本申请实施例提供的一种通信设备的硬件结构示意图；
图18是本申请实施例提供的一种网元的硬件结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚描述，显然，所描述的实施例是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员所获得的所有其他实施例，都属于本申请保护的范围。
本申请的说明书和权利要求书中的术语“第一”、“第二”等是用于区别类似的对象，而不用于描述特定的顺序或先后次序。应该理解这样使用的术语在适当情况下可以互换，以便本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施，且“第一”、“第二”所区别的对象通常为一类，并不限定对象的个数，例如第一对象可以是一个，也可以是多个。此外，说明书以及权利要求中“和/或”表示所连接对象的至少其中之一，字符“/”一般表示前后关联对象是一种“或”的关系。
值得指出的是，本申请实施例所描述的技术不限于长期演进型(Long Term Evolution，LTE)/LTE的演进(LTE-Advanced，LTE-A)系统，还可用于其他无线通信系统，诸如码分多址(Code Division Multiple Access，CDMA)、时分多址(Time Division Multiple Access，TDMA)、频分多址(Frequency Division Multiple Access，FDMA)、正交频分多址(Orthogonal Frequency Division Multiple Access，OFDMA)、单载波频分多址(Single-carrier Frequency Division Multiple Access，SC-FDMA)和其他系统。本申请实施例中的术语“系统”和“网络”常被可互换地使用，所描述的技术既可用于以上提及的系统和无线电技术，也可用于其他系统和无线电技术。以下描述出于示例目的描述了新空口(New Radio，NR)系统，并且在以下大部分描述中使用NR术语，但是这些技术也可应用于NR系统应用以外的应用，如第6代(6th Generation，6G)通信系统。
图1示出本申请实施例可应用的一种无线通信系统的框图。无线通信系统包括终端11和网络侧设备12。其中，终端11可以是手机、平板电脑(Tablet Personal Computer)、膝上型电脑(Laptop Computer)或称为笔记本电脑、个人数字助理(Personal Digital Assistant，PDA)、掌上电脑、上网本、超级移动个人计算机(ultra-mobile personal computer，UMPC)、移动上网装置(Mobile Internet Device，MID)、增强现实(augmented reality，AR)/虚拟现实(virtual reality，VR)设备、机器人、可穿戴式设备(Wearable Device)、车载设备(VUE)、行人终端(PUE)、智能家居(具有无线通信功能的家居设备，如冰箱、电视、洗衣机或者家具等)、游戏机、个人计算机(personal computer，PC)、柜员机或者自助机等终端侧设备，可穿戴式设备包括：智能手表、智能手环、智能耳机、智能眼镜、智能首饰(智能手镯、智能手链、智能戒指、智能项链、智能脚镯、智能脚链等)、智能腕带、智能服装等。需要说明的是，在本申请实施例并不限定终端11的具体类型。网络侧设备12可以包括接入网设备或核心网设备，其中，接入网设备12也可以称为无线接入网设备、无线接入网(Radio Access Network,RAN)、无线接入网功能或无线接入网单元。接入网设备12可以包括基站、WLAN接入点或WiFi节点等，基站可被称为节点B、演进节点B(eNB)、接入点、基收发机站(Base Transceiver Station，BTS)、无线电基站、无线电收发机、基本服务集(Basic Service Set，BSS)、扩展服务集(Extended Service Set，ESS)、家用B节点、家用演进型B节点、发送接收点(Transmitting Receiving Point，TRP)或所述领域中其他某个合适的术语，只要达到相同的技术效果，所述基站不限于特定技术词汇，需要说明的是，在本申请实施例中仅以NR系统中的基站为例进行介绍，并不限定基站的具体类型。核心网设备可以包含但不限于如下至少一项：核心网节点、核心网功能、移动管理实体(Mobility Management Entity，MME)、接入移动管理功能(Access and Mobility Management Function，AMF)、会话管理功能(Session Management Function，SMF)、用户平面功能(User Plane Function，UPF)、策略控制功能(Policy Control Function，PCF)、策略与计费规则功能单元(Policy and Charging Rules Function，PCRF)、边缘应用服务发现功能(Edge Application Server Discovery Function，EASDF)、统一数据管理(Unified Data Management，UDM)，统一数据仓储(Unified Data Repository，UDR)、归属用户服务器(Home Subscriber Server，HSS)、集中式网络配置(Centralized network configuration，CNC)、网络存储功能(Network Repository Function，NRF)，网络开放功能(Network Exposure Function，NEF)、本地NEF(Local NEF，或L-NEF)、绑定支持功能(Binding Support Function，BSF)、应用功能(Application Function，AF)等。需要说明的是，在本申请实施例中仅以NR系统中的核心网设备为例进行介绍，并不限定核心网设备的具体类型。
下面对本申请实施例提供的网元注册方法、模型确定方法、装置、网元、通信系统及存储介质中涉 及的一些概念和/或术语做一下解释说明。
1、横向联邦学习：
适用场景：横向联邦学习的本质是样本的联合，适用于参与者间业务形态相同但客户不同，即特征重叠多，用户重叠少时的场景，比如第5代(5th Generation，5G)通信系统不同城市间的不同用户(例如每一个UE，即样本不同)的同一服务(例如视频业务、语音业务，其他互联网公司越过运营商(Over The Top，OTT)业务。通过联合参与方的不同样本的相同数据特征，横向联邦使训练样本的数量增多，从而得到一个更好的模型。
如图2所示，Server A为协调者(coordinator)，负责向其他客户(clients)、成员发送联邦学习的任务、模型等信息，并收集clients反馈的模型、参数、梯度、变化率等信息，更新初始模型，并在之后下发更新完成的模型；图2为横向联邦学习的原理图；
其中，①为发送加密的梯度(Sending encrypted gradients)；
②为安全的聚合(Secure aggregation)；
③为发回模型更新(Sending back model updates)；
④为更新模型(Updating models)。
2、人工智能(Artificial Intelligence，AI)及AI模型：
人工智能目前在各个领域获得了广泛的应用。AI模型有多种算法实现方式，例如神经网络、决策树、支持向量机、贝叶斯分类器等。本申请以神经网络为例进行说明，但是并不限定AI模块的具体类型。
如图3所示的，一个神经网络的示意图。X1，X2…Xn等为输入值，Y为输出结果，一个个○代表一个个神经元也是进行运算的地方，结果会继续传入到下一层。这些众多神经元组成的一输入层、隐藏层、输出层就是一个神经网络。隐藏层的数量，每一层神经元的数量就是神经网络的“网络结构”。
其中，神经网络由神经元组成，如图4所示的，神经元的示意图。其中a1，a2，…aK(即上文的X1…Xn)为输入，w为权值(乘性系数)，b为偏置(加性系数)，σ(z)为激活函数，z为输出值。常见的激活函数包括Sigmoid、tanh、ReLU(Rectified Linear Unit，线性整流函数，修正线性单元)等。每一个神经元的参数信息和所用算法组合在一起就是整个网络的“参数信息”，也是AI模型文件中很重要的一部分。
在实际使用过程中，一个AI模型指的是一个包含网络结构和参数信息等元素的文件，经过训练的AI模型可被其框架平台直接再次使用，无需重复构建或者学习，直接进行判断，识别等智能化功能。
3、网络数据分析功能(Network Data Analytics Function，NWDAF)网元可被分解为两部分，NWDAF(AnLF)和NWDAF(MTLF)，前者为负责推理功能(Analytics Logical Function，AnLF)网元，后者为负责训练功能(Model Training Logical Function，MTLF)网元。
下面结合附图，通过一些实施例及其应用场景对本申请实施例提供的网元注册方法、模型确定方法进行详细地说明。
5G网络中有多个NWDAF-MTLF，比如NWDAF-MTLF(coordinator)、NWDAF-MTLF1、NWDAF-MTLF2，其中：
1)NWDAF-MTLF(server A，coordinator)负责江苏省全省数据分析，但无法拥有江苏全省数据(由于数据都分布存储在江苏各个地市)；
2)NWDAF-MTLF1(Database B1，participant 1)能收集苏州地域的数据，并在苏州本地训练AI model；
3)NWDAF-MTLF2(Database B2，participant 2)能收集徐州地域的数据，并在苏州本地训练AI model。
所以，如果为了获得基于江苏全省的数据训练出来的AI Model，需要使用横向联邦学习，即不要求NWDAF-MTLF0获得全省数据。
因此如何在通信网络中基于联邦学习所得模型进行联邦推理的流程和细节还不清楚，如哪些网元参与联邦推理、过程中如何交互信息、交互哪些必要的信息等问题不清楚。
现有网元注册发现技术中只有对网元的标识、服务和部分能力信息的需求，但是对其关于联邦学习能力的注册和发现暂时不在标准范围内。
本申请实施例中，NWDAF网元、基站、UE等不同域中可以参加联邦学习的设备需向NRF等负责能力储存的网元发起注册过程，指示其具备参加联邦学习的能力信息。联邦学习发起者(比如MTRL网元)发起联邦学习训练时，该联邦学习发起者向NRF等网元请求信息，以寻找合适的联邦学习client(比如NWDAF网元)。在联邦学习训练过程结束后，联邦学习发起者MTLF网元向AnLF网元发送模型和联邦学习结果的同时，可指出参与联邦学习的参与者信息。
具体的，一种方案，NWDAF等网元可以通过注册请求，将联邦学习能力信息注册到NRF网元中，以实现将NWDAF等网元注册到NRF网元中，使得其他网元能够通过NRF网元查找到能够进行联邦学习训练的网元，以解决NWDAF等网元如何将联邦学习能力信息注册到NRF网元，并能够被其他网元查找到的问题。
另一种方案，NRF网元可以通过接收NWDAF等网元发送的注册请求，将NWDAF等网元的联邦学习能力信息注册到NRF网元中，以实现将NWDAF等网元注册到NRF网元中，使得其他网元能够通过NRF网元查找到能够进行联邦学习训练的网元，以解决NWDAF等网元如何将联邦学习能力信息 注册到NRF网元，并能够被其他网元查找到的问题。
又一种方案，MTRL网元可以通过查找请求，向NRF网元查找能够进行联邦学习训练的目标网元，由于查找请求中包括第一信息，因此MTRL网元可以查找到与第一信息匹配的目标网元。从而与查找到的能够进行联邦学习训练的目标网元进行联邦学习训练。
本申请实施例提供一种网元注册方法，图5示出了本申请实施例提供的一种网元注册方法的流程图。如图5所示，本申请实施例提供的网元注册方法可以包括下述的步骤201和步骤202。
步骤201、第一网元向第二网元发送注册请求。
本申请实施例中，上述注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元。
步骤202、第二网元接收第一网元发送的注册请求。
本申请实施例中，上述第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。
可选地，本申请实施例中，上述第一网元可以为NWDAF网元、NWDAF containing MTLF网元，第二网元为负责能力储存的网元，该网元可以为NRF网元、UDM网元、数据收集应用功能(Data Collection Application Function，DCAF)网元。
可选地，本申请实施例中，上述第一网元可以发送Nnrf_NFManagement_NFRegister Register注册请求到第二网元，以请求将第一网元注册到第二网元中。
需要说明的是，上述第一网元支持的联邦学习训练的类型信息用于指示第一网元所支持的AI模型训练算法类型。
可选地，本申请实施例中，上述第一网元支持的联邦学习训练的类型信息包括以下至少一项：第一网元是否支持联邦学习训练的指示信息、横向联邦学习训练类型、纵向联邦学习训练类型、联邦学习服务器能力、联邦学习客户端能力。
需要说明的是，上述联邦学习服务器(server)能力用于指示第一网元是否有联邦学习服务器能力，或者用于指示第一网元是否支持作为联邦学习服务器。
可选地，本申请实施例中，第一网元可以作为联邦学习服务器，其具有聚合各个联邦学习客户端提供的本地模型训练信息生成全局模型(Aggregate model)，或协调联邦学习(Coordinate federated learning)过程的能力。
需要说明的是，上述联邦学习客户端(clients)能力用于指示第一网元是否有联邦学习客户端能力，或者用于指示第一网元是否支持作为联邦学习客户端。
可选地，本申请实施例中，第一网元可以作为联邦学习客户端，其具有参与到联邦学习中的能力，具有进行本地模型的训练并可以提供本地模型训练信息的能力。
需要说明的是，上述第一网元支持联邦学习训练的时间信息用于指示第一网元所支持的联邦学习训练进行时间，在该时间段内进行联邦学习训练可以有较好的表现，例如：晚上10点到第二天早上6点。
需要说明的是，上述第一网元所具有的元数据信息用于指示第一网元所能覆盖、获取、提供的数据信息。
可选地，本申请实施例中，上述第一网元所具有的元数据信息包括以下至少一项：输入数据类型、输出数据类型、数据量、数据范围。
需要说明的是，上述输入数据类型是指联邦学习训练的输入数据类型，上述输出数据类型是指联邦学习训练的输出数据类型。
需要说明的是，上述输入数据类型用于指示第一网元可采集的用于作为输入数据进行模型训练过程的数据类型，或者第一网元可获取的数据类型。
可选地，本申请实施例中，上述数据类型是指区域内的数据是否有明显特征，例如早上聚集，晚上散开。
可选地，本申请实施例中，上述数据类型可以为数据类别，例如：UE的位置信息、UE的时间信息、网元的负载信息、网络状态信息、网元资源信息等。
可选地，本申请实施例中，上述数据范围是指第一网元的服务范围。
可选地，本申请实施例中，上述数据范围包括以下至少一项：第一网元的服务区域、第一网元可采集数据的区域、第一网元可采集数据的对象。
可选地，本申请实施例中，上述第一网元可采集数据包括用于模型训练的元数据信息和训练数据。
可选地，本申请实施例中，上述第一网元可采集数据的区域可以是以下任一项：第一网元服务区域、第一网元服务区域下的子区域范围、更细粒度的可采集数据粒度的对象(例如：一个UE list)。
可选地，本申请实施例中，上述第一网元可采集数据的对象可以包括一个或多个特定的网元，或者一个或多个UE。
可选地，本申请实施例中，上述注册请求中还包括以下至少一项：第一网元的标识信息、第一网元支持的数据分析标识信息、第一网元支持的模型过滤器信息、第一网元的网元类型的信息，该网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型。
可选地，本申请实施例中，上述第一网元支持的数据分析标识信息可以为用户移动性轨迹(UE mobility)信息。
可选地，本申请实施例中，上述数据分析标识信息与联邦学习训练的类型信息对应。
可选地，本申请实施例中，上述数据分析标识信息与第一网元的联邦学习能力信息对应。
可选地，本申请实施例中，第一网元的联邦学习能力信息是和数据分析标识信息对应的，例如第一网元的联邦学习能力信息可以是和某一数据分析标识对应的，即指该第一网元对于该数据分析标识的任务有联邦学习能力。
可选地，本申请实施例中，数据分析标识信息与第一网元的联邦学习能力信息存在映射关系，该映射关系可以为一对一、多对一或一对多；其中，在映射关系为一对一时，针对不同的数据分析标识信息，第一网元的联邦学习能力信息也对应不同。
例如，第一网元支持10个数据分析标识信息，数据分析标识1-3对应支持横向联邦学习训练，数据分析标识4-6对应支持纵向联邦学习训练，数据分析标识7-10对应不支持联邦学习训练。
可选地，本申请实施例中，上述第一网元的标识信息可以为以下任一项：全限定域名(Fully Qualified Domain Name，FQDN)信息、IP地址信息。
可选地，本申请实施例中，全限定域名信息用于指示第一网元的位置以及连接第一网元。
需要说明的是，上述模型过滤器信息(Model filter information)也可以称为模型有效范围信息，该模型有效范围信息用于指示第一网元训练产生的模型的有效范围。
可选地，本申请实施例中，上述模型有效范围信息包括模型适用的切片、模型适用的外部数据网络(Data Network Name，DNN)、模型适用的区域、模型适用的时间等。
可选地，本申请实施例中，上述模型适用的时间可以为一个大的时间范围，例如：一年内；也可以为周期性的时间，例如：每天的9点-10点。
可选地，本申请实施例中，上述第一网元支持的模型过滤器信息包括第一网元经过联邦学习所产生的模型对应的模型过滤器信息。
可选地，本申请实施例中，上述第一网元的联邦学习能力信息还包括以下至少一项：
第一网元的联邦学习训练所基于的算法信息；
第一网元的联邦学习训练可达到的模型准确度信息；
第一网元的联邦学习训练的速度信息；
第一网元支持的模型描述方式信息；
第一网元支持的模型可共享信息；
第一网元的厂商信息；
第一网元支持的联邦学习模型的数量信息。
需要说明的是，上述第一网元的联邦学习训练所基于的算法信息为训练模型使用的具体算法，例如：线性回归、神经网络。
需要说明的是，上述第一网元的联邦学习训练可达到的模型准确度信息用于指示模型输出结果的准确程度。
需要说明的是，上述第一网元的联邦学习训练的速度信息包括第一网元基于联邦学习训练的模型到达该模型的目标准确度所需时长信息。
需要说明的是，上述第一网元支持的模型描述方式信息用于指示第一网元支持基于该模型描述方式信息对应的模型描述方式表示模型。
可选地，本申请实施例中，上述模型描述方式信息也可以称作模型描述方式要求信息，或者模型描述方式期望信息。
可选地，本申请实施例中，模型描述方式信息可以是开放神经网络交换(Open Neural Network Exchange，ONNX)等为代表的模型表达语言，或者TensorFlow，Pytorch等为代表的模型框架。
需要说明的是，上述第一网元支持的模型可共享信息用于指示该第一网元支持共享模型，其中，可共享是指可以进行互操作，或者可共享是指可以相互被理解，可共享是指可运行。
可选地，本申请实施例中，上述模型可共享信息也可以称作模型可共享要求信息，或者模型可共享期望信息。
具体地，在第一网元支持的模型可共享信息包括一个或多个厂商信息时，其他属于该厂商信息中的网元可以与该第一网元进行交互，例如获取第一网元训练的模型和模型信息等。
具体地，在第一网元支持的模型可共享信息包括一个或多个模型信息时，其他网元可以与该模型信息对应的第一网元所训练的模型进行交互，例如获取第一网元训练的模型和模型信息等。
需要说明的是，上述模型可共享信息也可以是基于数据分析任务粒度的，即，对于每一类数据分析任务，有不同的可共享信息(不同的厂商信息，模型信息等)。
需要说明的是，上述第一网元的厂商信息用于指示第一网元对应的厂商。
需要说明的是，上述第一网元的厂商信息用于后续第二网元判断第一网元基于联邦学习训练所产生的模型是否可以与其他网元共享。
例如：其他网元的厂商信息和第一网元的厂商信息相同，则其他网元可以共享该第一网元基于联邦学习训练所产生的模型。
可选地，本申请实施例中，上述第一网元的厂商信息可以用于第二网元选择能够参与联邦学习训练的网元。
需要说明的是，上述第一网元支持的联邦学习模型的数量信息用于指示针对每个数据分析标识，第一网元分别支持的基于联邦学习训练的模型的数量信息，如最大数量和最小数量。
可选地，本申请实施例中，上述注册请求中还包括以下至少一项：第一网元的类型信息、第一网元 支持的服务名称。
例如：第一网元为NWDAF网元，NWDAF网元支持的服务名称为Nnwdaf_AnalyticsInfo_Request。
可选地，本申请实施例中，上述第一网元的类型信息用于指示此次注册的是何种网元，例如：第一网元的类型信息为NWDAF网元类型，指示此次注册的网元是NWDAF网元。
可选地，本申请实施例中，在上述步骤202之后，还包括下述的步骤203和步骤204。
步骤203、第二网元向第一网元发送注册响应。
本申请实施例中，上述注册响应用于指示第一设备能力注册成功。
可选地，本申请实施例中，上述第二网元可以发送Nnrf_NFManagement_NFRegister response注册响应到第二网元，以通知第一网元注册成功。
步骤204、第一网元接收第二网元发送的注册响应。
本申请实施例提供一种网元注册方法，第一网元可以通过注册请求，将第一网元的联邦学习能力信息注册到第二网元中，以实现将第一网元注册到第二网元中，从而使得其他网元能够通过第二网元查找到能够进行联邦学习训练的第一网元，以解决第一网元如何将联邦学习能力信息注册到第二网元，并能够被其他网元查找到的问题。
可选地，本申请实施例中，结合图5，如图6所示，在上述步骤202之后，本申请实施例提供的网元注册方法还包括下述的步骤205和步骤206。
步骤205、第三网元向第二网元发送查找请求。
本申请实施例中，上述查找请求用于请求查找能够进行联邦学习训练的网元。
步骤206、第二网元接收第三网元发送的查找请求。
本申请实施例中，上述查找请求中包括第一信息，该第一信息包括以下至少一项：目标任务对应的数据分析标识信息、目标任务对应的联邦学习训练的类型信息、目标任务对应的联邦学习训练的时间信息、目标任务对应的联邦学习训练的网元所具有的元数据信息、网元类型要求信息、网元数量要求信息，该网元类型要求信息用于指示所需查找的能够进行联邦学习训练的网元对应的网元类型，该网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型，网元数量要求信息用于指示所需查找的能够进行联邦学习训练的网元的数量要求。
本申请实施例中，上述第三网元为模型训练逻辑功能(MTLF，Model Training Logical Function)网元，即模型训练网元，可以理解为发起联邦学习的网元。
可选地，本申请实施例中，上述第三网元可以发送Nnrf_NFDiscovery_Request查找请求到第二网元，以请求查找能够进行联邦学习训练的网元。
可选地，本申请实施例中，在第三网元不支持/不进行独立训练生成目标任务对应的模型信息的情况下，第三网元向第二网元发送查找请求。
可选地，本申请实施例中，上述第三网元不支持/不进行独立训练生成目标任务对应的模型信息的情况可以为第三网元数据量不够的情况，或者第三网元训练模型的准确率不达标的情况。
可选地，本申请实施例中，上述第三网元可以是具有联邦学习服务器能力的网元，该第三网元可以寻找有联邦学习客户端能力的网元，以进行联邦学习训练。
可选地，本申请实施例中，上述第三网元可以是具有联邦学习客户端能力的网元，该第三网元可以寻找有联邦学习服务器能力的网元，以请求具有联邦学习服务器能力的网元组织进行联邦学习训练。
需要说明的是，上述目标任务对应的联邦学习训练的类型信息用于指示获取支持目标任务对应的联邦学习训练的类型的网元的网元信息，例如：第三网元判断需要进行联邦学习训练，则指示获取支持联邦学习训练的网元的网元信息，该网元信息用于指示网元支持联邦学习训练；或者，第三网元判断需要进行横向联邦学习训练，则指示获取支持横向联邦学习训练的网元的网元信息，该网元信息用于指示网元支持横向联邦学习训练。
需要说明的是，上述目标任务对应的联邦学习训练的时间信息用于指示第三网元计划进行联邦学习训练的时间信息，以查找到支持此时间范围内进行联邦学习训练的网元，该时间信息是未来的一个时间段，例如：晚上10点到第二天早上6点。
需要说明的是，上述目标任务对应的联邦学习训练的网元所具有的元数据信息用于指示需求目标网元所能覆盖、获取、提供的数据信息。
可选地，本申请实施例中，上述目标任务对应的联邦学习训练的网元所具有的元数据信息包括以下至少一项：输入数据类型、输出数据类型、数据量、数据范围。
可选地，本申请实施例中，上述第一信息还包括第二信息，该第二信息用于请求获取多个所需查找的能够进行联邦学习训练的网元。
可选地，本申请实施例中，上述网元数量要求信息可以包括以下至少一项：最小数量要求信息，建议数量信息，最大数量要求信息。
可选地，本申请实施例中，上述第一信息还包括以下至少一项：
目标任务对应的联邦学习训练所基于的算法信息；
目标任务对应的联邦学习训练可达到的模型准确度信息；
目标任务对应的联邦学习训练的速度信息；
目标任务对应的联邦学习训练的网元支持的模型描述方式信息；
目标任务对应的联邦学习训练的网元支持的模型可共享信息；
目标任务对应的联邦学习训练的网元的厂商信息；
目标任务对应的联邦学习训练的网元支持的联邦学习模型的数量信息。
需要说明的是，上述目标任务对应的联邦学习训练所基于的算法信息用于指示训练模型使用的具体算法，例如：线性回归、神经网络。
需要说明的是，上述目标任务对应的联邦学习训练可达到的模型准确度信息用于指示模型输出结果需要达到的准确程度。
需要说明的是，上述目标任务对应的联邦学习训练的速度信息用于指示网元基于联邦学习训练的模型到达该模型的目标准确度所需时长信息。
需要说明的是，上述目标任务对应的联邦学习训练的网元支持的模型描述方式信息用于指示网元需要支持基于该模型描述方式信息对应的模型描述方式表示模型。
需要说明的是，上述目标任务对应的联邦学习训练的网元支持的模型可共享信息用于指示网元需要支持共享模型。
需要说明的是，上述目标任务对应的联邦学习训练的网元的厂商信息用于指示网元对应的厂商。
需要说明的是，上述目标任务对应的联邦学习训练的网元支持的联邦学习模型的数量信息用于指示针对每个数据分析标识，网元需要分别支持的基于联邦学习训练的模型的数量。
可选地，本申请实施例中，上述第一信息还包括以下至少一项：目标任务对应的联邦学习训练的网元的服务信息、目标任务对应的联邦学习训练的网元的类型信息、感兴趣区域信息。
可选地，本申请实施例中，若目标服务是数据分析标识信息的请求，则上述目标任务对应的联邦学习训练的网元的服务信息可以为Nnwdaf_AnalyticsInfo_Request。
可选地，本申请实施例中，若目标服务的网元为NWDAF网元，则上述目标任务对应的联邦学习训练的网元的类型信息可以为NWDAF网元类型信息。
可选地，本申请实施例中，上述感兴趣区域信息可以是TA(s)，cell(s)或其他表现形式，用于指示请求在感兴趣区域内发现参与联邦学习训练的网元。
可选地，本申请实施例中，结合图6，如图7所示，在上述步骤206之后，本申请实施例提供的网元注册方法还包括下述的步骤207。
步骤207、第二网元根据查找请求确定目标网元。
本申请实施例中，上述目标网元的联邦学习能力信息与第一信息匹配。
可以理解，第二网元将联邦学习能力信息与第一信息匹配的网元确定为目标网元(一个或多个网元)。
可选地，本申请实施例中，上述步骤207具体可以通过下述的步骤207a至步骤207f实现。
步骤207a、第二网元根据目标任务对应的数据分析标识信息，确定目标网元。
本申请实施例中，上述目标网元支持数据分析标识信息对应的联邦学习训练。
可以理解，第二网元将支持数据分析标识信息对应的联邦学习训练的网元确定为目标网元。
步骤207b、第二网元根据目标任务对应的联邦学习训练的类型，确定目标网元。
本申请实施例中，上述目标网元支持目标任务对应的联邦学习训练的类型。
可以理解，第二网元将支持目标任务对应的联邦学习训练的类型的网元确定为目标网元。
可选地，本申请实施例中，第二网元可以将支持目标任务对应的联邦学习训练的网元确定为目标网元。
可选地，本申请实施例中，第二网元可以将支持目标任务对应的联邦学习服务器能力的网元确定为目标网元。
可选地，本申请实施例中，第二网元可以将支持目标任务对应的联邦学习客户端能力的网元确定为目标网元。
步骤207c、第二网元根据目标任务对应的联邦学习训练的时间信息，确定目标网元。
本申请实施例中，上述目标网元支持在目标时间信息对应的时间下进行联邦学习训练，目标时间信息为目标任务对应的联邦学习训练的时间信息。
可以理解，第二网元将支持在目标时间信息对应的时间下进行联邦学习训练的网元确定为目标网元。
步骤207d、第二网元根据目标任务对应的联邦学习训练的网元所具有的元数据信息，确定目标网元。
本申请实施例中，上述目标网元支持目标任务对应的联邦学习训练的网元所具有的元数据信息。
可以理解，第二网元将支持目标任务对应的联邦学习训练的网元确定为目标网元。
步骤207e、第二网元根据网元数量要求信息，确定目标网元。
本申请实施例中，上述目标网元的数量满足网元数量要求信息。
可以理解，在第二网元查找到的能够进行联邦学习训练的网元的数量，满足网元数量要求信息指示的网元数量时，可以将符合数量要求的网元确定为目标网元。
步骤207f、第二网元根据网元类型要求信息，确定目标网元。
本申请实施例中，上述目标网元的网元类型满足网元类型要求信息。
可以理解，第二网元将满足网元类型要求信息的网元确定为目标网元。
可选地，本申请实施例中，第二网元根据目标任务对应的联邦学习训练所基于的算法信息，确定目 标网元。
可以理解，第二网元将支持目标任务对应的联邦学习训练所基于的算法的网元确定为目标网元。
可选地，本申请实施例中，第二网元根据目标任务对应的联邦学习训练可达到的模型准确度信息，确定目标网元。
可以理解，第二网元将能够达到目标任务对应的联邦学习训练的模型准确度的网元确定为目标网元。
可选地，本申请实施例中，第二网元根据目标任务对应的联邦学习训练的速度信息，确定目标网元。
可以理解，第二网元将支持目标任务对应的联邦学习训练的速度的网元确定为目标网元；
可选地，本申请实施例中，第二网元根据目标任务对应的联邦学习训练的网元支持的模型描述方式信息，确定目标网元。
可以理解，第二网元将支持上述模型描述方式信息(即目标任务对应的联邦学习训练的网元支持的模型描述方式信息)对应的模型描述方式的网元确定为目标网元。
可选地，本申请实施例中，第二网元根据目标任务对应的联邦学习训练的网元支持的模型可共享信息，确定目标网元。
可以理解，第二网元将支持共享模型的网元确定为目标网元。
可选地，本申请实施例中，第二网元根据目标任务对应的联邦学习训练的网元的厂商信息，确定目标网元。
可以理解，第二网元将与厂商信息对应的网元确定为目标网元。
可选地，本申请实施例中，第二网元根据目标任务对应的联邦学习训练的网元支持的联邦学习模型的数量信息，确定目标网元。
可以理解，第二网元将支持上述联邦学习模型的数量信息(即目标任务对应的联邦学习训练的网元支持的联邦学习模型的数量信息)的网元，确定为目标网元。
可选地，本申请实施例中，在第二网元根据网元数量要求信息，确定目标网元的情况下，本申请实施例提供的网元注册方法还包括下述的步骤A1。
步骤A1、在第二网元查找到的能够进行联邦学习训练的网元的数量小于网元数量要求信息指示的网元的数量的情况下，第二网元将查找到的能够进行联邦学习训练的网元确定为目标网元。
可以理解，上述查找到的能够进行联邦学习训练的网元为满足第三网元的其他要求信息(例如：网元类型、联邦学习训练的时间等)，但不满足网元数量要求信息的网元。
可选地，本申请实施例中，在第二网元查找到的能够进行联邦学习训练的网元的数量等于网元数量要求信息指示的网元的数量的情况下，第二网元将查找到的能够进行联邦学习训练的网元确定为目标网元。
可以理解，上述查找到的能够进行联邦学习训练的网元为满足第三网元的要求信息的网元。
可选地，本申请实施例中，在第二网元查找到的能够进行联邦学习训练的网元的数量大于网元数量要求信息指示的网元的数量的情况下，第二网元将查找到的能够进行联邦学习训练的网元中的部分网元确定为目标网元。
可选地，本申请实施例中，第二网元可以按照预设的内部逻辑对查找到的所有能够进行联邦学习训练的网元进行排序，然后按照排序结果从中选择满足网元数量要求信息的网元，作为目标网元。
可选地，本申请实施例中，结合图7，如图8所示，在上述步骤207之后，本申请实施例提供的网元注册方法还包括下述的步骤208和步骤209。
步骤208、第二网元向第三网元发送查找响应。
本申请实施例中，上述查找响应中包括目标网元的标识信息或地址信息。
步骤209、第三网元接收第二网元发送的查找响应。
可选地，本申请实施例中，上述目标网元的标识信息可以为以下任一项：FQDN信息、IP地址信息。
可选地，本申请实施例中，上述第二网元可以发送Nnrf_NFDiscovery_Request查找响应到第三网元，以发送能够进行联邦学习训练的网元的网元信息。
可选地，本申请实施例中，上述查找响应中还包括第二信息，第二信息包括以下至少一项：目标网元支持的联邦学习训练的类型信息、目标网元支持联邦学习训练的时间信息、目标网元所具有的元数据信息。
本申请实施例中，上述目标网元支持的联邦学习训练的类型信息用于指示目标网元所支持的AI模型训练算法类型。
本申请实施例中，上述目标网元支持联邦学习训练的时间信息用于指示目标网元所支持的联邦学习训练进行时间，在该时间段内进行联邦学习训练可以有较好的表现，例如：晚上10点到第二天早上6点。
本申请实施例中，上述目标网元所具有的元数据信息用于指示目标网元所能覆盖、获取、提供的数据信息。
可选地，本申请实施例中，上述目标网元所具有的元数据信息包括以下至少一项：输入数据类型、输出数据类型、数据量、数据范围。
可选地，本申请实施例中，上述查找响应中还包括目标网元的类型信息。
可选地，本申请实施例中，上述目标网元的类型信息用于指示目标网元是何种网元，例如：目标网元的类型信息为NWDAF网元类型，指示目标网元是NWDAF网元。
可选地，本申请实施例中，在第二网元查找到的能够进行联邦学习训练的网元的数量小于网元数量要求信息指示的网元的数量的情况下，查找响应中还包括第三信息，该第三信息包括以下至少一项：
第一指示信息，该第一指示信息用于指示目标网元的数量小于网元数量要求信息指示的网元的数量；
目标网元的数量信息；
第四信息，该第四信息用于建议第三网元延后查找能够进行联邦学习训练的网元，该第四信息包括：延后查找能够进行联邦学习训练的网元的时间信息。
可选地，本申请实施例中，上述时间信息可以为一个具体的时间。
例如：上述第四信息可以包括时间信息“半小时”，以使得第三网元收到第三信息时，可以根据该时间信息，在半小时后再次查找能够进行联邦学习训练的网元。
可以理解，在第二网元查找到的能够进行联邦学习训练的网元的数量小于网元数量要求信息指示的网元的数量时，一方面，第二网元可以将查找到的能够进行联邦学习训练的网元确定为目标网元，并在查找响应中携带目标网元的标识信息或地址信息；另一方面，第二网元可以在查找响应中携带第三信息，以告知第三网元以下至少一项：目标网元的数量不满足网元数量要求信息指示的数量、目标网元的数量信息、建议第三网元可以延后查找能够进行联邦学习训练的网元。
可选地，本申请实施例中，在第二网元查找到的能够进行联邦学习训练的网元的数量等于网元数量要求信息指示的网元的数量的情况下，查找响应中还可以包括第五信息，该第五信息用于指示目标网元的数量满足网元数量要求信息指示的网元的数量。
可以理解，在第二网元查找到的能够进行联邦学习训练的网元的数量等于网元数量要求信息指示的网元的数量时，一方面，第二网元可以将查找到的能够进行联邦学习训练的网元确定为目标网元，并在查找响应中携带目标网元的标识信息或地址信息；另一方面，第二网元可以在查找响应中携带第五信息，以告知第三网元，目标网元的数量满足网元数量要求信息指示的数量。
可选地，本申请实施例中，在第二网元查找到的能够进行联邦学习训练的网元的数量大于网元数量要求信息指示的网元的数量的情况下，查找响应中还可以包括第六信息，该第六信息包括以下至少一项：
第三指示信息，该第三指示信息用于指示查找到的能够进行联邦学习训练的网元的数量大于网元数量要求信息指示的网元的数量；
查找到的能够进行联邦学习训练的网元的相关信息；
查找到的能够进行联邦学习训练的网元的数量信息。
可以理解，在第二网元查找到的能够进行联邦学习训练的网元的数量大于网元数量要求信息指示的网元的数量时，一方面，第二网元可以从能够进行联邦学习训练的网元中选择满足网元数量要求信息的部分网元作为目标网元，并在查找响应中携带目标网元的标识信息或地址信息；另一方面，第二网元可以在查找响应中携带第六信息，以告知第三网元以下至少一项：第二网元查找到的所有能够进行联邦学习训练的网元的相关信息、第二网元查找到的所有能够进行联邦学习训练的网元的数量信息、第二网元查找到的能够进行联邦学习训练的网元的数量超过网元数量要求信息指示的数量。
可选地，本申请实施例中，第三网元接收到第三信息之后，可以通过与目标网元进行联邦学习训练，得到目标任务对应的模型信息；也可以不与第二网元确定的目标网元进行联邦学习训练，即停止进行联邦学习训练；也可以基于第四信息或者内部逻辑，延后重新发起查找请求，以重新查找能够进行联邦学习训练的网元，进行联邦学习。
可选地，本申请实施例中，第三网元接收到第六信息之后，可以从目标网元中选择全部或部分网元进行联邦学习训练，以得到目标任务对应的模型信息；也可以基于内部逻辑从第二网元查找到的所有能够进行联邦学习训练的网元中选择部分网元进行联邦学习训练，以得到目标任务对应的模型信息。
可选地，本申请实施例中，在上述步骤209之后，本申请实施例提供的网元注册方法还包括下述的步骤209a。
步骤209a、第三网元从目标网元中选择全部或部分网元进行联邦学习训练。
可选地，本申请实施例中，上述全部或部分网元可以作为联邦学习训练的客户端或联邦学习训练的服务器；具体地，第三网元可以根据网元类型和网元类型要求信息从目标网元中选择进行联邦学习训练的客户端或进行联邦学习训练的服务器。
可选地，本申请实施例中，上述第三网元可以基于目标网元的联邦学习能力，或者目标网元的注册信息，或者第三网元的内部逻辑等从目标网元中选择全部或部分网元进行联邦学习训练。
本申请实施例提供一种网元注册方法，第二网元可以通过接收第三网元发送的查找请求，以根据查找请求确定能够进行联邦学习训练的目标网元，然后可以向第三网元发送查找响应，以使得第三网元确定能够进行联邦学习训练的目标网元，从而与能够进行联邦学习训练的目标网元进行联邦学习训练，以得到目标任务对应的模型信息。
需要说明的是，本申请实施例提供的网元注册方法，执行主体还可以为网元注册装置，或者，该网元注册装置中用于执行网元注册方法的控制模块。
本申请实施例提供一种模型确定方法，图9示出了本申请实施例提供的一种模型确定方法的流程图。如图9所示，本申请实施例提供的模型确定方法可以包括下述的步骤301和步骤302。
步骤301、第三网元向第二网元发送查找请求。
本申请实施例中，上述查找请求用于请求查找能够进行联邦学习训练的网元。
步骤302、第二网元接收第三网元发送的查找请求。
可选地，本申请实施例中，在第三网元不支持/不进行独立训练生成目标任务对应的模型信息的情况下，第三网元向第二网元发送查找请求。
可选地，本申请实施例中，上述第三网元不支持/不进行独立训练生成目标任务对应的模型信息的情况可以为第三网元数据量不够的情况，或者第三网元训练模型的准确率不达标的情况。
本申请实施例中，上述第三网元为模型训练逻辑功能(MTLF，Model Training Logical Function)网元，即模型训练网元，可以理解为发起联邦学习的网元。
可选地，本申请实施例中，上述第三网元可以发送Nnrf_NFDiscovery_Request查找请求到第二网元，以请求查找能够进行联邦学习训练的网元。
本申请实施例中，上述查找请求中包括第一信息，第一信息包括以下至少一项：目标任务对应的数据分析标识信息、目标任务对应的联邦学习训练的类型信息、目标任务对应的联邦学习训练的时间信息、目标任务对应的联邦学习训练的网元所具有的元数据信息、网元类型要求信息、网元数量要求信息，该网元类型要求信息用于指示所需查找的能够进行联邦学习训练的网元对应的网元类型，该网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型，网元数量要求信息用于指示所需查找的能够进行联邦学习训练的网元的数量要求。
可选地，本申请实施例中，上述目标任务对应的联邦学习训练的类型信息用于指示获取支持目标任务对应的联邦学习训练的类型的网元的网元信息，例如：第三网元判断需要进行联邦学习训练，则指示获取支持联邦学习训练的网元的网元信息，该网元信息用于指示网元支持联邦学习训练；或者，第三网元判断需要进行横向联邦学习训练，则指示获取支持横向联邦学习训练的网元的网元信息，该网元信息用于指示网元支持横向联邦学习训练。
可选地，本申请实施例中，上述目标任务对应的联邦学习训练的时间信息用于指示第三网元计划进行联邦学习训练的时间信息，以查找到支持此时间范围内进行联邦学习训练的网元，该时间信息是未来的一个时间段，例如：晚上10点到第二天早上6点。
可选地，本申请实施例中，上述目标任务对应的联邦学习训练的网元所具有的元数据信息用于指示需求目标网元所能覆盖、获取、提供的数据信息。
可选地，本申请实施例中，上述目标任务对应的联邦学习训练的网元所具有的元数据信息包括以下至少一项：输入数据类型、输出数据类型、数据量、数据范围。
可选地，本申请实施例中，上述输入数据类型是指联邦学习训练的输入数据类型，上述输出数据类型是指联邦学习训练的输出数据类型。
可选地，本申请实施例中，上述数据类型是指联邦学习训练的输入数据、输出数据是否有明显特征，例如早上聚集，晚上散开。
可选地，本申请实施例中，上述数据范围是指目标网元的服务范围。
可选地，本申请实施例中，上述数据范围包括以下至少一项：目标网元的服务区域、目标网元可采集数据的区域、目标网元可采集数据的对象。
可选地，本申请实施例中，上述目标网元可采集数据包括用于模型训练的元数据信息和训练数据。
可选地，本申请实施例中，上述目标网元可采集数据的区域可以是以下任一项：目标网元服务区域、目标网元服务区域下的子区域范围、更细粒度的可采集数据粒度的对象(例如：一个UE list)。
可选地，本申请实施例中，上述第一信息还包括以下至少一项：
目标任务对应的联邦学习训练所基于的算法信息；
目标任务对应的联邦学习训练可达到的模型准确度信息；
目标任务对应的联邦学习训练的速度信息；
目标任务对应的联邦学习训练的网元支持的模型描述方式信息；
目标任务对应的联邦学习训练的网元支持的模型可共享信息；
目标任务对应的联邦学习训练的网元的厂商信息；
目标任务对应的联邦学习训练的网元支持的联邦学习模型的数量信息。
可选地，本申请实施例中，上述目标任务对应的联邦学习训练所基于的算法信息用于指示训练模型使用的具体算法，例如：线性回归、神经网络。
可选地，本申请实施例中，上述目标任务对应的联邦学习训练可达到的模型准确度信息用于指示模型输出结果需要达到的准确程度。
可选地，本申请实施例中，上述目标任务对应的联邦学习训练的速度信息用于指示网元基于联邦学习训练的模型到达该模型的目标准确度所需时长信息。
可选地，本申请实施例中，上述第一信息还包括以下至少一项：目标任务对应的联邦学习训练的网元的服务信息、目标任务对应的联邦学习训练的网元的类型信息、感兴趣区域信息。
可选地，本申请实施例中，若目标服务是数据分析标识信息的请求，则上述目标任务对应的联邦学习训练的网元的服务信息可以为Nnwdaf_AnalyticsInfo_Request。
可选地，本申请实施例中，若目标服务的网元为NWDAF网元，则上述目标任务对应的联邦学习训练的网元的类型信息可以为NWDAF网元类型信息。
可选地，本申请实施例中，上述感兴趣区域信息可以是TA(s)，cell(s)或其他表现形式，用于指示请求在感兴趣区域内发现参与联邦学习训练的网元。
本申请实施例提供一种模型确定方法，第三网元可以通过查找请求，向第二网元查找能够进行联邦学习训练的网元，由于查找请求中包括第一信息，因此第三网元可以查找到与第一信息匹配的目标网元。从而与查找到的能够进行联邦学习训练的目标网元进行联邦学习训练。
可选地，本申请实施例中，结合图9，如图10所示，在上述步骤302之后，本申请实施例提供的模型确定方法还包括下述的步骤303至步骤305。
步骤303、第二网元向第三网元发送查找响应。
本申请实施例中，上述查找响应中包括目标网元的标识信息或地址信息，目标网元为支持联邦学习训练的网元。
可选地，本申请实施例中，上述目标网元的标识信息可以为以下任一项：FQDN信息、IP地址信息。
步骤304、第三网元接收第二网元发送的查找响应。
可选地，本申请实施例中，上述第二网元可以发送Nnrf_NFDiscovery_Request查找响应到第三网元，以发送能够进行联邦学习训练的网元的网元信息。
可选地，本申请实施例中，上述查找响应中还包括第二信息，第二信息包括以下至少一项：目标网元支持的联邦学习训练的类型信息、目标网元支持联邦学习训练的时间信息、目标网元所具有的元数据信息。
可选地，本申请实施例中，上述目标网元支持的联邦学习训练的类型信息用于指示目标网元所支持的AI模型训练算法类型。
可选地，本申请实施例中，上述目标网元支持联邦学习训练的时间信息用于指示目标网元所支持的联邦学习训练进行时间，在该时间段内进行联邦学习训练可以有较好的表现，例如：晚上10点到第二天早上6点。
可选地，本申请实施例中，上述目标网元所具有的元数据信息用于指示目标网元所能覆盖、获取、提供的数据信息。
可选地，本申请实施例中，上述目标网元所具有的元数据信息包括以下至少一项：输入数据类型、输出数据类型、数据量、数据范围。
可选地，本申请实施例中，上述查找响应中还包括目标网元的类型信息。
可选地，本申请实施例中，上述目标网元的类型信息用于指示目标网元是何种网元，例如：目标网元的类型信息为NWDAF网元类型，指示目标网元是NWDAF网元。
步骤305、第三网元通过与目标网元进行联邦学习训练，得到目标任务对应的模型信息。
可选地，本申请实施例中，第三网元可以根据上述接收的查找响应中包括的信息确定并寻找参与联邦学习训练的目标网元，以与寻找到的目标网元进行联邦学习。
可选地，本申请实施例中，结合图10，如图11所示，在上述步骤301之前，本申请实施例提供的模型确定方法还包括下述的步骤306和步骤307。
步骤306、第四网元向第三网元发送模型请求。
本申请实施例中，上述模型请求用于请求目标任务对应的模型信息。
步骤307、第三网元接收第四网元发送的模型请求。
可选地，本申请实施例中，上述第四网元为模型推理网元，该网元可以为AnLF网元、NWDAF网元、NWDAF containing AnLF网元。
可选地，本申请实施例中，上述第四网元可以发送Nnwdaf_MLMoldelInfo_Request模型请求到第三网元，以请求第三网元反馈符合任务的模型信息。
可选地，本申请实施例中，上述模型请求中包括以下至少一项：目标任务对应的数据分析标识信息、目标任务对应的限定信息。
可选地，本申请实施例中，上述目标任务对应的数据分析标识信息可以为用户移动性轨迹(UE mobility)信息。
可选地，本申请实施例中，上述目标任务对应的限定信息用于指示目标任务的具体限定条件，例如：限定目标任务所需要的时间、位置等信息。
可选地，本申请实施例中，上述模型请求中还包括以下至少一项：目标网元的指示信息、报告限定信息。
可选地，本申请实施例中，上述目标网元的指示信息用于指示目标任务的目标网元，例如：用户移动性轨迹信息中，可以规定目标网元的SUPI等身份标识信息。
可选地，本申请实施例中，上述报告限定信息用于指示目标任务所需要汇报的信息、格式等，例如：排列方式为升序。
可选地，本申请实施例中，结合图11，如图12所示，在上述步骤305之后，本申请实施例提供的模型确定方法还包括下述的步骤308和步骤309。
步骤308、第三网元向第四网元发送模型响应。
本申请实施例中，上述模型响应中包括目标任务对应的模型信息，模型信息用于第四网元进行模型推理。
步骤309、第四网元接收第三网元发送的模型响应。
可选地，本申请实施例中，上述目标任务对应的模型信息包括以下至少一项：目标信息、联邦学习训练相关信息。
可选地，本申请实施例中，上述目标信息包括以下至少一项：目标任务对应的模型描述信息、目标任务对应的模型文件、目标任务对应的模型存储地址信息。
可选地，本申请实施例中，上述目标任务对应的模型文件中包括以下至少一项：生成目标任务对应的模型信息的完整的网络结构、生成目标任务对应的模型信息的参数信息。
可选地，本申请实施例中，上述联邦学习训练相关信息包括以下至少一项：第二指示信息、参与联邦学习训练的网元的标识信息、参与联邦学习训练的网元的能力信息。
本申请实施例中，上述第二指示信息用于指示目标任务对应的模型信息为通过联邦学习训练得到的模型信息。
可选地，本申请实施例中，上述参与联邦学习训练的网元的标识信息可以为网元的身份、地址信息，例如：全限定域名信息、IP地址信息等。
可选地，本申请实施例中，第三网元可以将参与联邦学习训练的网元的标识信息一起发送给第四网元，以供第四网元在后续进行联邦学习训练的推理。
可选地，本申请实施例中，上述目标任务对应的模型信息还包括以下至少一项：目标任务对应的数据分析标识信息、目标任务对应的限定信息。
可选地，本申请实施例中，上述目标任务对应的限定信息中包括感兴趣区域的信息或目标终端的信息。本申请实施例提供的模型确定方法还可以包括下述的步骤401。
步骤401、在第三网元无法获取感兴趣区域或目标终端对应的目标任务的训练数据的情况下，第三网元执行针对目标任务的联邦学习训练。
需要说明的是，上述感兴趣区域是基于联邦学习训练的模型需要/待预测信息的区域。
例如：通过模型预测某一个区域明天的天气。
需要说明的是，上述目标终端是基于联邦学习训练的模型需要/待预测信息的终端。
例如：通过模型预测目标终端明天的行程。
可选地，本申请实施例中，上述目标任务对应的限定信息中包括感兴趣区域的信息或目标终端的信息。本申请实施例提供的模型确定方法还可以包括下述的步骤402。
步骤402、第三网元根据感兴趣区域的信息或目标终端的信息确定目标任务对应的联邦学习训练的网元所具有的元数据信息。
本申请实施例中，上述元数据信息包括联邦学习训练的网元的数据范围。
本申请实施例提供一种模型确定方法，第三网元可以接收第四网元发送的模型请求，请求目标任务对应的模型信息，然后第三网元可以通过查找请求，向第二网元查找能够进行联邦学习训练的网元，由于查找请求中包括第一信息，因此第三网元可以查找到与第一信息匹配的目标网元。通过与查找到的目标网元进行联邦学习训练，以得到目标任务对应的模型信息，从而提高了模型传递的成功率。
需要说明的是，本申请实施例提供的模型确定方法，执行主体还可以为模型确定装置，或者，该模型确定装置中用于执行模型确定方法的控制模块。
下面通过具体的实施方式，对本申请实施例提供的网元注册及模型确定方法的交互过程进行详细地说明。
如图13所示，本申请实施例提供的方法包括下述的步骤21至步骤32。
步骤21、不同域中的智能化设备(例如，核心网侧的NWDAF网元，基站，UE，第三方应用服务器等)向NRF等能力储存设备(例如，NRF网元，UDM网元，DCAF网元等)发送能力注册消息，进行能力注册。
本申请实施例中，NWDAF网元可通过(Nnrf_NFManagement_NFRegister Register)注册。
步骤22、NRF网元储存该NWDAF网元的信息。
步骤23、NRF网元发送注册响应消息。
本申请实施例中，NRF网元可通过(Nnrf_NFManagement_NFRegister response)回应消息，通知NWDAF网元注册成功。
这些步骤与现有技术类似，不同点在于：
步骤21中，NWDAF网元向NRF网元发送能力注册消息时，除了自身标识信息、支持的analytic ID等信息外，还发送“支持的训练类型信息”，“支持联邦学习时间”和“元数据信息Meta data”等。
注册所必须的信息有：
1、NF type，网元类型；指此次注册的是何种网元，如此方案中，NF type＝NWDAF type等。
2、NF instance ID,FQDN or IP address of NF，网元实例标识信息；指此次注册的网元标识信息，例如，其FQDN信息(Fully Qualified Domain Name，全限定域名，用于指示此网元的位置和连接此网元)或者IP地址信息(另一种标识信息)。
3、Names of supported NF services(if applicable)，网元所支持的服务名称；如NWDAF网元，有服务名为Nnwdaf_AnalyticsInfo_Request等。
另外，网元注册信息中还需包含以下至少一种：
4、支持的训练类型信息；指该网元所支持的AI模型训练算法类型，如“联邦学习”，“深度学 习”等此类信息。
5、支持联邦学习时间；指该网元所支持的联邦学习进行时间，在该时间段内进行联邦学习可以有较好的表现，如“晚上10点到第二天早上6点”。
6、元数据信息Meta data；指该网元所能覆盖、获取、提供的数据信息，包括数据类型、数据特征、数据量等信息。
7、Analytics ID，数据分析ID；如“UE mobility”(用户移动性轨迹)等
8、模型训练所用算法；训练模型使用的具体算法，如线性回归，神经网络等。
9、模型训练可达模型准确度；模型输出结果的准确程度。
10、模型训练的速度；用于指示模型训练到达特定准确度所需要的时长。
步骤24、任务消费者(consumer)向模型推理网元AnLF(或NWDAF，NWDAF containing AnLF)发送数据分析请求消息。
本申请实施例中，任务消费者(consumer)可通过(Nnwdaf_AnalyticsInfo_Request或者Nnwdaf_AnalyticsSubscription_Subscribe)发送数据分析请求消息。
步骤25、AnLF接收到任务请求后向MTLF发送模型请求。请求MTLF反馈符合任务的模型。
本申请实施例中，AnLF可通过(Nnwdaf_MLMoldelInfo_Request)发送模型请求，该请求中，包含任务描述信息，如：
1、analytic ID，数据分析ID；如“UE mobility”(用户移动性轨迹)等。
2、filter info，任务限定信息；指示此任务的具体限定条件，如限定任务所需要的时间、位置等信息。
3、目标网元/UE等信息；指示任务的目标网元或者UE，如UE mobility中，可规定目标UE的SUPI等身份标识信息。
4、reporting info，报告限定信息；指定任务所需汇报的信息、格式等，如排列方式为升序。
步骤26、MTLF根据步骤25中的任务描述信息进行判断是否可以独立训练生成符合要求的模型，若不可独立训练生成，则可发起联邦学习。
本申请实施例中，不能独立训练生成的情况可能是数据集数量不够，数据特征区别不够等无法使准确率达到目标准确率等情况。
步骤27、模型训练网元MTLF等发起联邦训练的网元设备向NRF等储存能力网元发送寻找可以进行联邦学习训练的网元等设备的请求。
本申请实施例中，模型训练网元MTLF可通过(Nnrf_NFDiscovery_Request)发送查找请求。
这些步骤与现有技术类似，不同点在于：
MTLF向NRF发送寻找参与联邦学习的网元的请求，其需包括：
1、target NF service Name(s),目标网元服务信息；如目标服务是数据分析信息的请求，则该信息可为Nnwdaf_AnalyticsInfo_Request。
2、NF type of the target NF,目标网元的类型；如目标服务的网元为NWDAF，则该信息可为NWDAF type。
3、感兴趣区域，area of interest；可以是TA(s)，cell(s)或其他表现形式，用于指示请求在感兴趣区域内发现参与联邦学习的网元。
MTLF向NRF发送请求时，还需包含“训练类型”，“训练条件限定信息”等。
4、训练类型；指希望获取到网元设备所支持的训练类型，如在此实施例中，网元希望进行联邦学习，则希望获取到的网元信息支持联邦学习的训练，则该信息可为指示网元支持联邦学习的信息，如“联邦学习”等。
5、训练条件限定信息；指对网元设备等的筛选信息，对网元设备等的范围的进一步缩小以选出更适合需要的网元，如“联邦学习时间信息”，“所需数据信息”等。包括以下至少一项：
支持联邦学习时间；指计划进行联邦学习的目标时间信息，目标选出能在此时间范围内可进行联邦学习的网元设备。该时间信息将是未来的一个时间段，如“晚上10点到第二天早上6点”。
所需数据信息；指希望目标网元所能覆盖、获取、提供的数据信息，包括数据类型、数据特征、数据量等信息。
步骤28、NRF将符合步骤27中检索信息的网元等设备信息返回给MTLF。
本申请实施例中，NRF等能力存储网元向联邦学习发起者返回符合步骤26中训练条件限定信息的网元设备信息，可通过(Nnrf_NFDiscovery_Request)返回响应消息。
其中需包括：
1、网元设备标识信息；指示目标网元、设备的FQDN、IP地址等。
2、网元设备的类型信息；如NWDAF。
3、网元设备的能力信息；包括以下至少一项：
训练类型；指希望获取到网元设备所支持的训练类型，如“联邦学习”等。
支持联邦学习时间；指该网元所支持的联邦学习进行时间，在该时间段内进行联邦学习可以有较好的表现，如“晚上10点到第二天早上6点”。
元数据信息；指该网元所能覆盖、获取、提供的数据信息，包括数据类型、数据特征、数据量等信息。
步骤29、MTLF根据步骤28中返回的信息确定并寻找参与训练的网元等设备，MTLF与寻找到的网元等设备进行联邦学习。
步骤30、MTLF训练结束后，向AnLF发送反馈模型，其中包括针对任务的描述信息。
1、analytic ID；任务类型的指示信息。
2、任务限定信息；进一步描述任务需求的信息，如限定时间、位置等信息。
3、模型等；
还可包括关于联邦学习成员的标识信息和能力信息，包括以下至少一项：
4、联邦学习信息；指示此次训练使用联邦学习完成
5、参与联邦学习成员标识信息；可以是成员的身份、地址信息，如FQDN或IP地址等。MTLF将参与联邦学习的网元等设备信息一起发送给AnLF，供AnLF在后续进行联邦学习的推理。
6、参与联邦学习成员的能力信息；包括支持联邦学习的时间信息和设备自身的数据信息等。
步骤31、AnLF接收到反馈的模型等信息后，进行模型推理。
本申请实施例中，AnLF可以根据获得的联邦学习信息中的参与联邦学习的网元等设备信息进行合作推理。
步骤32、AnLF根据推理结果向consumer反馈任务报告。
需要说明的是，针对上述步骤21至步骤32中相关说明和有益效果，可以参见上述实施例中的描述，此处不再赘述。
图14示出了本申请实施例中涉及的网元注册装置的一种可能的结构示意图。如图14所示，网元注册装置80可以包括：发送模块81。
其中，发送模块81，用于向第二网元发送注册请求，该注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元，第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。
本申请实施例提供一种网元注册装置，网元注册装置可以通过注册请求，将第一网元的联邦学习能力信息注册到第二网元中，以实现将第一网元注册到第二网元中，从而使得其他网元能够通过第二网元查找到能够进行联邦学习训练的第一网元，以解决第一网元如何将联邦学习能力信息注册到第二网元，并能够被其他网元查找到的问题。
在一种可能的实现方式中，上述第一网元支持的联邦学习训练的类型信息包括以下至少一项：第一网元是否支持联邦学习训练的指示信息、横向联邦学习训练类型、纵向联邦学习训练类型、联邦学习服务器能力、联邦学习客户端能力。
在一种可能的实现方式中，上述第一网元所具有的元数据信息包括以下至少一项：输入数据类型、输出数据类型、数据量、数据范围。
在一种可能的实现方式中，上述数据范围包括以下至少一项：第一网元的服务区域、第一网元可采集数据的区域、第一网元可采集数据的对象。
在一种可能的实现方式中，上述注册请求中还包括以下至少一项：第一网元的标识信息、第一网元支持的数据分析标识信息、第一网元支持的模型过滤器信息、第一网元的网元类型的信息，网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型。
在一种可能的实现方式中，上述数据分析标识信息与联邦学习训练的类型信息对应。
在一种可能的实现方式中，上述数据分析标识信息与第一网元的联邦学习能力信息对应。
在一种可能的实现方式中，上述模型过滤器信息包括第一网元经过联邦学习所产生的模型对应的模型过滤器信息。
在一种可能的实现方式中，上述第一网元的联邦学习能力信息还包括以下至少一项：第一网元的联邦学习训练所基于的算法信息；第一网元的联邦学习训练可达到的模型准确度信息；第一网元的联邦学习训练的速度信息；第一网元支持的模型描述方式信息；第一网元支持的模型可共享信息；第一网元的厂商信息；第一网元支持的联邦学习模型的数量信息。
在一种可能的实现方式中，上述第一网元的联邦学习训练的速度信息包括第一网元基于联邦学习训练的模型到达该模型的目标准确度所需时长信息。
本申请实施例提供的网元注册装置能够实现上述方法实施例中第一网元实现的各个过程，并达到相同的技术效果，为避免重复，这里不再赘述。
图15示出了本申请实施例中涉及的网元注册装置的一种可能的结构示意图。如图15所示，网元注册装置90可以包括：接收模块91。
其中，接收模块91，用于接收第一网元发送的注册请求，该注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元，第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。
本申请实施例提供一种网元注册装置，网元注册装置可以通过接收第一网元发送的注册请求，将第一网元的联邦学习能力信息注册到第二网元中，以实现将第一网元注册到第二网元中，从而使得其他网元查找到能够进行联邦学习训练的第一网元，以解决第一网元如何将联邦学习能力信息注册到第二网元，并能够被其他网元查找到的问题。
在一种可能的实现方式中，上述第一网元的联邦学习能力信息还包括以下至少一项：
第一网元的联邦学习训练所基于的算法信息；第一网元的联邦学习训练可达到的模型准确度信息； 第一网元的联邦学习训练的速度信息；第一网元支持的模型描述方式信息；第一网元支持的模型可共享信息；第一网元的厂商信息；第一网元支持的联邦学习模型的数量信息。
在一种可能的实现方式中，上述接收模块91，还用于接收第三网元发送的查找请求，该查找请求用于请求查找能够进行联邦学习训练的网元，查找请求中包括第一信息，第一信息包括以下至少一项：目标任务对应的数据分析标识信息、目标任务对应的联邦学习训练的类型信息、目标任务对应的联邦学习训练的时间信息、目标任务对应的联邦学习训练的网元所具有的元数据信息、网元类型要求信息、网元数量要求信息，该网元类型要求信息用于指示所需查找的能够进行联邦学习训练的网元对应的网元类型，该网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型，网元数量要求信息用于指示所需查找的能够进行联邦学习训练的网元的数量要求。
在一种可能的实现方式中，上述第一信息还包括以下至少一项：目标任务对应的联邦学习训练所基于的算法信息；目标任务对应的联邦学习训练可达到的模型准确度信息；目标任务对应的联邦学习训练的速度信息；目标任务对应的联邦学习训练的网元支持的模型描述方式信息；目标任务对应的联邦学习训练的网元支持的模型可共享信息；目标任务对应的联邦学习训练的网元的厂商信息；目标任务对应的联邦学习训练的网元支持的联邦学习模型的数量信息。
在一种可能的实现方式中，本申请实施例提供的网元注册装置90还包括：确定模块。确定模块，用于在接收模块91接收第三网元发送的查找请求之后，根据查找请求确定目标网元，该目标网元的联邦学习能力信息与第一信息匹配。
在一种可能的实现方式中，上述确定模块，具体用于根据目标任务对应的数据分析标识信息，确定目标网元，其中，目标网元支持数据分析标识信息对应的联邦学习训练；根据目标任务对应的联邦学习训练的类型，确定目标网元，其中，目标网元支持目标任务对应的联邦学习训练的类型；根据目标任务对应的联邦学习训练的时间信息，确定目标网元，其中，目标网元支持在目标时间信息对应的时间下进行联邦学习训练，目标时间信息为目标任务对应的联邦学习训练的时间信息；根据目标任务对应的联邦学习训练的网元所具有的元数据信息，确定目标网元，其中，目标网元支持目标任务对应的联邦学习训练的网元所具有的元数据信息；根据网元数量要求信息，确定目标网元，其中，目标网元的数量满足网元数量要求信息；根据网元类型要求信息，确定目标网元，其中，目标网元的网元类型满足网元类型要求信息。
在一种可能的实现方式中，在第二网元根据网元数量要求信息，确定目标网元的情况下，上述确定模块，用于在第二网元查找到的能够进行联邦学习训练的网元的数量小于网元数量要求信息指示的网元的数量的情况下，将查找到的能够进行联邦学习训练的网元确定为目标网元。
在一种可能的实现方式中，本申请实施例提供的网元注册装置90还包括：发送模块。发送模块，用于在确定模块根据查找请求确定目标网元之后，向第三网元发送查找响应，该查找响应中包括目标网元的标识信息或地址信息。
在一种可能的实现方式中，上述查找响应中还包括第二信息，该第二信息包括以下至少一项：目标网元支持的联邦学习训练的类型信息、目标网元支持联邦学习训练的时间信息、目标网元所具有的元数据信息。
在一种可能的实现方式中，在第二网元查找到的能够进行联邦学习训练的网元的数量小于网元数量要求信息指示的网元的数量的情况下，查找响应中还包括第三信息，该第三信息包括以下至少一项：第一指示信息，该第一指示信息用于指示目标网元的数量小于网元数量要求信息指示的网元的数量；目标网元的数量信息；第四信息，该第四信息用于建议第三网元延后查找能够进行联邦学习训练的网元，该第四信息包括：延后查找能够进行联邦学习训练的网元的时间信息。
本申请实施例提供的网元注册装置能够实现上述方法实施例中第二网元实现的各个过程，并达到相同的技术效果，为避免重复，这里不再赘述。
图16示出了本申请实施例中涉及的模型确定装置的一种可能的结构示意图。如图16所示，模型确定装置100可以包括：发送模块101。
其中，发送模块101，用于向第二网元发送查找请求，该查找请求用于请求查找能够进行联邦学习训练的网元，查找请求中包括第一信息，该第一信息包括以下至少一项：目标任务对应的数据分析标识信息、目标任务对应的联邦学习训练的类型信息、目标任务对应的联邦学习训练的时间信息、目标任务对应的联邦学习训练的网元所具有的元数据信息、网元类型要求信息、网元数量要求信息，该网元类型要求信息用于指示所需查找的能够进行联邦学习训练的网元对应的网元类型，该网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型，网元数量要求信息用于指示所需查找的能够进行联邦学习训练的网元的数量要求。
本申请实施例提供一种模型确定装置，模型确定装置可以通过查找请求，向第二网元查找能够进行联邦学习训练的网元，由于查找请求中包括第一信息，因此第三网元可以查找到与第一信息匹配的目标网元。从而与查找到的能够进行联邦学习训练的目标网元进行联邦学习训练。
在一种可能的实现方式中，上述目标任务对应的联邦学习训练的网元所具有的元数据信息包括以下至少一项：输入数据类型、输出数据类型、数据量、数据范围。
在一种可能的实现方式中，上述第一信息还包括以下至少一项：目标任务对应的联邦学习训练所基于的算法信息；目标任务对应的联邦学习训练可达到的模型准确度信息；目标任务对应的联邦学习训练的速度信息；目标任务对应的联邦学习训练的网元支持的模型描述方式信息；目标任务对应的联邦学习 训练的网元支持的模型可共享信息；目标任务对应的联邦学习训练的网元的厂商信息；目标任务对应的联邦学习训练的网元支持的联邦学习模型的数量信息。
在一种可能的实现方式中，本申请实施例提供的模型确定装置100还包括：接收模块和获取模块。接收模块，用于在发送模块101向第二网元发送查找请求之后，接收第二网元发送的查找响应，该查找响应中包括目标网元的标识信息或地址信息，该目标网元为支持联邦学习训练的网元；获取模块，用于通过与目标网元进行联邦学习训练，得到目标任务对应的模型信息。
在一种可能的实现方式中，上述查找响应中还包括第二信息，该第二信息包括以下至少一项：目标网元支持的联邦学习训练的类型信息、目标网元支持联邦学习训练的时间信息、目标网元所具有的元数据信息。
在一种可能的实现方式中，在第二网元查找到的能够进行联邦学习训练的网元的数量小于网元数量要求信息指示的网元的数量的情况下，查找响应中还包括第三信息，该第三信息包括以下至少一项：第一指示信息，该第一指示信息用于指示目标网元的数量小于网元数量要求信息指示的网元的数量；目标网元的数量信息；第四信息，该第四信息用于建议第三网元延后查找能够进行联邦学习训练的网元，该第四信息包括：延后查找能够进行联邦学习训练的网元的时间信息。
在一种可能的实现方式中，申请实施例提供的模型确定装置100还包括：选择模块；选择模块，用于在接收模块接收第二网元发送的查找响应之后，从目标网元中选择全部或部分网元进行联邦学习训练。
在一种可能的实现方式中，上述接收模块还用于在发送模块101向第二网元发送查找请求之前，接收第四网元发送的模型请求，该模型请求用于请求目标任务对应的模型信息，模型请求中包括以下至少一项：目标任务对应的数据分析标识信息、目标任务对应的限定信息。
在一种可能的实现方式中，上述发送模块101，还用于在获取模块通过与目标网元进行联邦学习训练，得到目标任务对应的模型信息之后，向第四网元发送模型响应，该模型响应中包括目标任务对应的模型信息，模型信息用于第四网元进行模型推理。
在一种可能的实现方式中，本申请实施例提供的模型确定装置100还包括：执行模块。目标任务对应的限定信息中包括感兴趣区域的信息或目标终端的信息；执行模块用于在第三网元无法获取感兴趣区域或目标终端对应的目标任务的训练数据的情况下，执行针对目标任务的联邦学习训练。
在一种可能的实现方式中，本申请实施例提供的模型确定装置100还包括：确定模块。目标任务对应的限定信息中包括感兴趣区域的信息或目标终端的信息；确定模块，用于根据感兴趣区域的信息或目标终端的信息确定目标任务对应的联邦学习训练的网元所具有的元数据信息，该元数据信息包括联邦学习训练的网元的数据范围。
在一种可能的实现方式中，上述目标任务对应的模型信息包括以下至少一项：目标信息、联邦学习训练相关信息，目标信息包括以下至少一项：目标任务对应的模型描述信息、目标任务对应的模型文件、目标任务对应的模型存储地址信息。
在一种可能的实现方式中，上述联邦学习训练相关信息包括以下至少一项：第二指示信息、参与联邦学习训练的网元的标识信息、参与联邦学习训练的网元的能力信息，第二指示信息用于指示目标任务对应的模型信息为通过联邦学习训练得到的模型信息。
本申请实施例提供的模型确定装置能够实现上述方法实施例中第三网元实现的各个过程，并达到相同的技术效果，为避免重复，这里不再赘述。
可选地，如图17所示，本申请实施例还提供一种通信设备5000，包括处理器5001和存储器5002，存储器5002上存储有可在所述处理器5001上运行的程序或指令，例如，该通信设备5000为网元时，该程序或指令被处理器5001执行时实现上述第一网元侧方法实施例的各个步骤，且能达到相同的技术效果，或实现上述第二网元侧方法实施例的各个步骤，且能达到相同的技术效果，或实现上述第三网元侧方法实施例的各个步骤，且能达到相同的技术效果，为避免重复，这里不再赘述。
本申请实施例还提供一种网元，包括处理器和通信接口，通信接口用于向第二网元发送注册请求，该注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元，第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。该网元实施例与上述第一网元侧方法实施例对应，上述方法实施例的各个实施过程和实现方式均可适用于该网元实施例中，且能达到相同的技术效果。
本申请实施例还提供一种网元，包括处理器和通信接口，通信接口用于接收第一网元发送的注册请求，该注册请求用于请求将第一网元的联邦学习能力信息注册到第二网元，第一网元的联邦学习能力信息包括以下至少一项：第一网元支持的联邦学习训练的类型信息、第一网元支持联邦学习训练的时间信息、第一网元所具有的元数据信息。该网元实施例与上述第二网元侧方法实施例对应，上述方法实施例的各个实施过程和实现方式均可适用于该网元实施例中，且能达到相同的技术效果。
本申请实施例还提供一种网元，包括处理器和通信接口，通信接口用于向第二网元发送查找请求，该查找请求用于请求查找能够进行联邦学习训练的网元，查找请求中包括第一信息，该第一信息包括以下至少一项：目标任务对应的数据分析标识信息、目标任务对应的联邦学习训练的类型信息、目标任务对应的联邦学习训练的时间信息、目标任务对应的联邦学习训练的网元所具有的元数据信息、网元类型要求信息、网元数量要求信息，该网元类型要求信息用于指示所需查找的能够进行联邦学习训练的网元对应的网元类型，该网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型，网元数量 要求信息用于指示所需查找的能够进行联邦学习训练的网元的数量要求。该网元实施例与上述第三网元侧方法实施例对应，上述方法实施例的各个实施过程和实现方式均可适用于该网元实施例中，且能达到相同的技术效果。
具体地，图18为实现本申请实施例的一种网元的硬件结构示意图，该网元为第一网元，或者该网元为第二网元，或者该网元为第三网元。
如图18所示，网元1200包括：处理器1201、网络接口1202和存储器1203。其中，网络接口1202例如为通用公共无线接口。
具体地，本申请实施例的网元1200还包括：存储在存储器1203上并可在处理器1201上运行的指令或程序，处理器1201调用存储器1203中的指令或程序执行上述各模块执行的方法，并达到相同的技术效果，为避免重复，故不在此赘述。
本申请实施例提供的网元能够实现上述方法实施例中第一网元、第二网元和第三网元实现的各个过程，并达到相同的技术效果，为避免重复，这里不再赘述。
本申请实施例还提供一种可读存储介质，所述可读存储介质上存储有程序或指令，该程序或指令被处理器执行时实现上述方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
其中，所述处理器为上述实施例中所述的通信设备中的处理器。所述可读存储介质，包括计算机可读存储介质，如计算机只读存储器ROM、随机存取存储器RAM、磁碟或者光盘等。
本申请实施例另提供了一种芯片，所述芯片包括处理器和通信接口，所述通信接口和所述处理器耦合，所述处理器用于运行程序或指令，实现上述方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
应理解，本申请实施例提到的芯片还可以称为系统级芯片，系统芯片，芯片系统或片上系统芯片等。
本申请实施例另提供了一种计算机程序/程序产品，所述计算机程序/程序产品被存储在存储介质中，所述计算机程序/程序产品被至少一个处理器执行以实现上述方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
本申请实施例还提供了一种通信系统，包括：第一网元、第二网元和第三网元，所述第一网元可用于执行如上所述的网元注册方法的步骤，所述第二网元可用于执行如上所述的网元注册方法的步骤，所述第三网元可用于执行如上所述的模型确定方法的步骤。
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者装置不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者装置所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、方法、物品或者装置中还存在另外的相同要素。此外，需要指出的是，本申请实施方式中的方法和装置的范围不限按示出或讨论的顺序来执行功能，还可包括根据所涉及的功能按基本同时的方式或按相反的顺序来执行功能，例如，可以按不同于所描述的次序来执行所描述的方法，并且还可以添加、省去、或组合各种步骤。另外，参照某些示例所描述的特征可在其他示例中被组合。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以计算机软件产品的形式体现出来，该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本申请各个实施例所述的方法。
上面结合附图对本申请的实施例进行了描述，但是本申请并不局限于上述的具体实施方式，上述的具体实施方式仅仅是示意性的，而不是限制性的，本领域的普通技术人员在本申请的启示下，在不脱离本申请宗旨和权利要求所保护的范围情况下，还可做出很多形式，均属于本申请的保护之内。

Claims (46)

  1. 一种网元注册方法，所述方法包括：
    第一网元向第二网元发送注册请求，所述注册请求用于请求将所述第一网元的联邦学习能力信息注册到所述第二网元，所述第一网元的联邦学习能力信息包括以下至少一项：所述第一网元支持的联邦学习训练的类型信息、所述第一网元支持联邦学习训练的时间信息、所述第一网元所具有的元数据信息。
  2. 根据权利要求1所述的方法，其中，所述第一网元支持的联邦学习训练的类型信息包括以下至少一项：所述第一网元是否支持联邦学习训练的指示信息、横向联邦学习训练类型、纵向联邦学习训练类型、联邦学习服务器能力、联邦学习客户端能力。
  3. 根据权利要求1所述的方法，其中，所述第一网元所具有的元数据信息包括以下至少一项：输入数据类型、输出数据类型、数据量、数据范围。
  4. 根据权利要求3所述的方法，其中，所述数据范围包括以下至少一项：所述第一网元的服务区域、所述第一网元可采集数据的区域、所述第一网元可采集数据的对象。
  5. 根据权利要求1至4中任一项所述的方法，其中，所述注册请求中还包括以下至少一项：所述第一网元的标识信息、所述第一网元支持的数据分析标识信息、所述第一网元支持的模型过滤器信息、所述第一网元的网元类型的信息，所述网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型。
  6. 根据权利要求5所述的方法，其中，所述数据分析标识信息与所述联邦学习训练的类型信息对应。
  7. 根据权利要求5所述的方法，其中，所述数据分析标识信息与所述第一网元的联邦学习能力信息对应。
  8. 根据权利要求5所述的方法，其中，所述模型过滤器信息包括所述第一网元经过联邦学习所产生的模型对应的模型过滤器信息。
  9. 根据权利要求1所述的方法，其中，所述第一网元的联邦学习能力信息还包括以下至少一项：
    所述第一网元的联邦学习训练所基于的算法信息；
    所述第一网元的联邦学习训练可达到的模型准确度信息；
    所述第一网元的联邦学习训练的速度信息；
    所述第一网元支持的模型描述方式信息；
    所述第一网元支持的模型可共享信息；
    所述第一网元的厂商信息；
    所述第一网元支持的联邦学习模型的数量信息。
  10. 根据权利要求9所述的方法，其中，所述第一网元的联邦学习训练的速度信息包括所述第一网元基于联邦学习训练的模型到达所述模型的目标准确度所需时长信息。
  11. 一种网元注册方法，所述方法包括：
    第二网元接收第一网元发送的注册请求，所述注册请求用于请求将所述第一网元的联邦学习能力信息注册到所述第二网元，所述第一网元的联邦学习能力信息包括以下至少一项：所述第一网元支持的联邦学习训练的类型信息、所述第一网元支持联邦学习训练的时间信息、所述第一网元所具有的元数据信息。
  12. 根据权利要求11所述的方法，其中，所述第一网元的联邦学习能力信息还包括以下至少一项：
    所述第一网元的联邦学习训练所基于的算法信息；
    所述第一网元的联邦学习训练可达到的模型准确度信息；
    所述第一网元的联邦学习训练的速度信息；
    所述第一网元支持的模型描述方式信息；
    所述第一网元支持的模型可共享信息；
    所述第一网元的厂商信息；
    所述第一网元支持的联邦学习模型的数量信息。
  13. 根据权利要求11所述的方法，其中，所述第二网元接收第一网元发送的注册请求之后，所述方法还包括：
    所述第二网元接收第三网元发送的查找请求，所述查找请求用于请求查找能够进行联邦学习训练的网元，所述查找请求中包括第一信息，所述第一信息包括以下至少一项：目标任务对应的数据分析标识信息、所述目标任务对应的联邦学习训练的类型信息、所述目标任务对应的联邦学习训练的时间信息、所述目标任务对应的联邦学习训练的网元所具有的元数据信息、网元类型要求信息、网元数量要求信息，所述网元类型要求信息用于指示所需查找的能够进行联邦学习训练的网元对应的网元类型，所述网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型，所述网元数量要求信息用于指示所需查找的能够进行联邦学习训练的网元的数量要求。
  14. 根据权利要求13所述的方法，其中，所述第一信息还包括以下至少一项：
    所述目标任务对应的联邦学习训练所基于的算法信息；
    所述目标任务对应的联邦学习训练可达到的模型准确度信息；
    所述目标任务对应的联邦学习训练的速度信息；
    所述目标任务对应的联邦学习训练的网元支持的模型描述方式信息；
    所述目标任务对应的联邦学习训练的网元支持的模型可共享信息；
    所述目标任务对应的联邦学习训练的网元的厂商信息；
    所述目标任务对应的联邦学习训练的网元支持的联邦学习模型的数量信息。
  15. 根据权利要求13或14所述的方法，其中，所述第二网元接收第三网元发送的查找请求之后，所述方法还包括：
    所述第二网元根据所述查找请求确定目标网元，所述目标网元的联邦学习能力信息与所述第一信息匹配。
  16. 根据权利要求15所述的方法，其中，所述第二网元根据所述查找请求确定目标网元，包括以下至少一项：
    所述第二网元根据所述目标任务对应的数据分析标识信息，确定所述目标网元，其中，所述目标网元支持所述数据分析标识信息对应的联邦学习训练；
    所述第二网元根据所述目标任务对应的联邦学习训练的类型，确定所述目标网元，其中，所述目标网元支持所述目标任务对应的联邦学习训练的类型；
    所述第二网元根据所述目标任务对应的联邦学习训练的时间信息，确定所述目标网元，其中，所述目标网元支持在所述目标时间信息对应的时间下进行联邦学习训练，所述目标时间信息为所述目标任务对应的联邦学习训练的时间信息；
    所述第二网元根据所述目标任务对应的联邦学习训练的网元所具有的元数据信息，确定所述目标网元，其中，所述目标网元支持所述目标任务对应的联邦学习训练的网元所具有的元数据信息；
    所述第二网元根据所述网元数量要求信息，确定所述目标网元，其中，所述目标网元的数量满足所述网元数量要求信息；
    所述第二网元根据所述网元类型要求信息，确定所述目标网元，其中，所述目标网元的网元类型满足所述网元类型要求信息。
  17. 根据权利要求16所述的方法，其中，在所述第二网元根据所述网元数量要求信息，确定所述目标网元的情况下，所述方法还包括：
    在所述第二网元查找到的能够进行联邦学习训练的网元的数量小于所述网元数量要求信息指示的网元的数量的情况下，所述第二网元将所述查找到的能够进行联邦学习训练的网元确定为所述目标网元。
  18. 根据权利要求15所述的方法，其中，所述第二网元根据所述查找请求确定目标网元之后，所述方法还包括：
    所述第二网元向所述第三网元发送查找响应，所述查找响应中包括所述目标网元的标识信息或地址信息。
  19. 根据权利要求18所述的方法，其中，所述查找响应中还包括第二信息，所述第二信息包括以下至少一项：
    所述目标网元支持的联邦学习训练的类型信息、所述目标网元支持联邦学习训练的时间信息、所述目标网元所具有的元数据信息。
  20. 根据权利要求18所述的方法，其中，在所述第二网元查找到的能够进行联邦学习训练的网元的数量小于所述网元数量要求信息指示的网元的数量的情况下，所述查找响应中还包括第三信息，所述第三信息包括以下至少一项：
    第一指示信息，所述第一指示信息用于指示所述目标网元的数量小于所述网元数量要求信息指示的网元的数量；
    所述目标网元的数量信息；
    第四信息，所述第四信息用于建议所述第三网元延后查找能够进行联邦学习训练的网元，所述第四信息包括：延后查找能够进行联邦学习训练的网元的时间信息。
  21. 一种模型确定方法，所述方法包括：
    第三网元向第二网元发送查找请求，所述查找请求用于请求查找能够进行联邦学习训练的网元，所述查找请求中包括第一信息，所述第一信息包括以下至少一项：目标任务对应的数据分析标识信息、所述目标任务对应的联邦学习训练的类型信息、所述目标任务对应的联邦学习训练的时间信息、所述目标任务对应的联邦学习训练的网元所具有的元数据信息、网元类型要求信息、网元数量要求信息，所述网元类型要求信息用于指示所需查找的能够进行联邦学习训练的网元对应的网元类型，所述网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型，所述网元数量要求信息用于指示所需查找的能够进行联邦学习训练的网元的数量要求。
  22. 根据权利要求21所述的方法，其中，所述目标任务对应的联邦学习训练的网元所具有的元数据信息包括以下至少一项：输入数据类型、输出数据类型、数据量、数据范围。
  23. 根据权利要求21所述的方法，其中，所述第一信息还包括以下至少一项：
    所述目标任务对应的联邦学习训练所基于的算法信息；
    所述目标任务对应的联邦学习训练可达到的模型准确度信息；
    所述目标任务对应的联邦学习训练的速度信息；
    所述目标任务对应的联邦学习训练的网元支持的模型描述方式信息；
    所述目标任务对应的联邦学习训练的网元支持的模型可共享信息；
    所述目标任务对应的联邦学习训练的网元的厂商信息；
    所述目标任务对应的联邦学习训练的网元支持的联邦学习模型的数量信息。
  24. 根据权利要求21至23中任一项所述的方法，其中，所述第三网元向第二网元发送查找请求之后，所述方法还包括：
    所述第三网元接收所述第二网元发送的查找响应，所述查找响应中包括目标网元的标识信息或地址信息，所述目标网元为支持联邦学习训练的网元；所述第三网元通过与所述目标网元进行联邦学习训练，得到所述目标任务对应的模型信息。
  25. 根据权利要求24所述的方法，其中，所述查找响应中还包括第二信息，所述第二信息包括以下至少一项：
    所述目标网元支持的联邦学习训练的类型信息、所述目标网元支持联邦学习训练的时间信息、所述目标网元所具有的元数据信息。
  26. 根据权利要求24所述的方法，其中，在所述第二网元查找到的能够进行联邦学习训练的网元的数量小于所述网元数量要求信息指示的网元的数量的情况下，所述查找响应中还包括第三信息，所述第三信息包括以下至少一项：
    第一指示信息，所述第一指示信息用于指示所述目标网元的数量小于所述网元数量要求信息指示的网元的数量；
    所述目标网元的数量信息；
    第四信息，所述第四信息用于建议所述第三网元延后查找能够进行联邦学习训练的网元，所述第四信息包括：延后查找能够进行联邦学习训练的网元的时间信息。
  27. 根据权利要求24所述的方法，其中，所述第三网元接收所述第二网元发送的查找响应之后，所述方法还包括：
    所述第三网元从所述目标网元中选择全部或部分网元进行联邦学习训练。
  28. 根据权利要求21所述的方法，其中，所述第三网元向第二网元发送查找请求之前，所述方法还包括：
    所述第三网元接收第四网元发送的模型请求，所述模型请求用于请求所述目标任务对应的模型信息，所述模型请求中包括以下至少一项：所述目标任务对应的数据分析标识信息、所述目标任务对应的限定信息。
  29. 根据权利要求24所述的方法，其中，所述第三网元通过与所述目标网元进行联邦学习训练，得到所述目标任务对应的模型信息之后，所述方法还包括：
    所述第三网元向所述第四网元发送模型响应，所述模型响应中包括所述目标任务对应的模型信息，所述模型信息用于所述第四网元进行模型推理。
  30. 根据权利要求28所述的方法，其中，所述目标任务对应的限定信息中包括感兴趣区域的信息或目标终端的信息；所述方法还包括：
    在所述第三网元无法获取所述感兴趣区域或所述目标终端对应的所述目标任务的训练数据的情况下，所述第三网元执行针对所述目标任务的联邦学习训练。
  31. 根据权利要求28所述的方法，其中，所述目标任务对应的限定信息中包括感兴趣区域的信息或目标终端的信息；所述方法还包括：
    所述第三网元根据所述感兴趣区域的信息或所述目标终端的信息确定所述目标任务对应的联邦学习训练的网元所具有的元数据信息，所述元数据信息包括所述联邦学习训练的网元的数据范围。
  32. 根据权利要求28所述的方法，其中，所述目标任务对应的模型信息包括以下至少一项：目标信息、联邦学习训练相关信息，所述目标信息包括以下至少一项：所述目标任务对应的模型描述信息、所述目标任务对应的模型文件、所述目标任务对应的模型存储地址信息。
  33. 根据权利要求32所述的方法，其中，所述联邦学习训练相关信息包括以下至少一项：第二指示信息、参与联邦学习训练的网元的标识信息、参与联邦学习训练的网元的能力信息，所述第二指示信息用于指示所述目标任务对应的模型信息为通过联邦学习训练得到的模型信息。
  34. 一种网元注册装置，应用于第一网元，所述装置包括：发送模块；
    所述发送模块，用于向第二网元发送注册请求，所述注册请求用于请求将所述第一网元的联邦学习能力信息注册到所述第二网元，所述第一网元的联邦学习能力信息包括以下至少一项：所述第一网元支持的联邦学习训练的类型信息、所述第一网元支持联邦学习训练的时间信息、所述第一网元所具有的元数据信息。
  35. 一种网元注册装置，应用于第二网元，所述装置包括：接收模块；
    所述接收模块，用于接收第一网元发送的注册请求，所述注册请求用于请求将所述第一网元 的联邦学习能力信息注册到所述第二网元，所述第一网元的联邦学习能力信息包括以下至少一项：所述第一网元支持的联邦学习训练的类型信息、所述第一网元支持联邦学习训练的时间信息、所述第一网元所具有的元数据信息。
  36. 一种模型确定装置，所述装置包括：发送模块；
    所述发送模块，用于向第二网元发送查找请求，所述查找请求用于请求查找能够进行联邦学习训练的网元，所述查找请求中包括第一信息，所述第一信息包括以下至少一项：目标任务对应的数据分析标识信息、所述目标任务对应的联邦学习训练的类型信息、所述目标任务对应的联邦学习训练的时间信息、所述目标任务对应的联邦学习训练的网元所具有的元数据信息、网元类型要求信息、网元数量要求信息，所述网元类型要求信息用于指示所需查找的能够进行联邦学习训练的网元对应的网元类型，所述网元类型包括联邦学习服务器网元类型和/或联邦学习客户端网元类型，所述网元数量要求信息用于指示所需查找的能够进行联邦学习训练的网元的数量要求。
  37. 一种网元，包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如权利要求1至10中任一项所述的网元注册方法的步骤。
  38. 一种网元，包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如权利要求11至20中任一项所述的网元注册方法的步骤。
  39. 一种网元，包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如权利要求21至33中任一项所述的模型确定方法的步骤。
  40. 一种通信系统，所述通信系统包括如权利要求34所述的网元注册装置、如权利要求35所述的网元注册装置以及如权利要求36所述的模型确定装置；或者，
    所述通信系统包括如权利要求37至39所述的网元。
  41. 一种可读存储介质，所述可读存储介质上存储程序或指令，所述程序或指令被处理器执行时实现如权利要求1至10中任一项所述的网元注册方法的步骤，或者实现如权利要求11至20中任一项所述的网元注册方法的步骤，或者实现如权利要求21至33中任一项所述的模型确定方法的步骤。
  42. 一种芯片，所述芯片包括处理器和通信接口，所述通信接口和所述处理器耦合，所述处理器用于运行程序或指令，实现如权利要求1至10中任一项所述的网元注册方法，或者实现如权利要求11至20中任一项所述的网元注册方法，或者实现如权利要求21至33中任一项所述的模型确定方法。
  43. 一种计算机程序产品，所述程序产品被至少一个处理器执行以实现如权利要求1至10中任一项所述的网元注册方法，或者实现如权利要求11至20中任一项所述的网元注册方法，或者实现如权利要求21至33中任一项所述的模型确定方法。
  44. 一种网元，包括所述网元被配置成用于执行如权利要求1至10中任一项所述的网元注册方法。
  45. 一种网元，包括所述网元被配置成用于执行如权利要求11至20中任一项所述的网元注册方法。
  46. 一种网元，包括所述网元被配置成用于执行如权利要求21至33中任一项所述的模型确定方法。
PCT/CN2023/084355 2022-03-28 2023-03-28 网元注册方法、模型确定方法、装置、网元、通信系统及存储介质 WO2023185822A1 (zh)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN202210317250 2022-03-28
CN202210317250.2 2022-03-28
CN202211468901 2022-11-22
CN202211468901.4 2022-11-22

Publications (1)

Publication Number Publication Date
WO2023185822A1 true WO2023185822A1 (zh) 2023-10-05

Family

ID=88199197

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/084355 WO2023185822A1 (zh) 2022-03-28 2023-03-28 网元注册方法、模型确定方法、装置、网元、通信系统及存储介质

Country Status (1)

Country Link
WO (1) WO2023185822A1 (zh)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021179176A1 (en) * 2020-03-10 2021-09-16 Asiainfo Technologies (China) , Inc. Federated learning in telecom communication system
CN113573331A (zh) * 2020-04-29 2021-10-29 华为技术有限公司 一种通信方法、装置及系统
CN114079902A (zh) * 2020-08-13 2022-02-22 Oppo广东移动通信有限公司 联邦学习的方法和装置

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021179176A1 (en) * 2020-03-10 2021-09-16 Asiainfo Technologies (China) , Inc. Federated learning in telecom communication system
CN113573331A (zh) * 2020-04-29 2021-10-29 华为技术有限公司 一种通信方法、装置及系统
CN114079902A (zh) * 2020-08-13 2022-02-22 Oppo广东移动通信有限公司 联邦学习的方法和装置

Similar Documents

Publication Publication Date Title
US20230144062A1 (en) Method for computational-power sharing and related device
CN104486129B (zh) 分布式环境下保障应用服务质量的方法及系统
CN115152191B (zh) 移动网络中的分析生成和使用
US11843516B2 (en) Federated learning in telecom communication system
WO2022001941A1 (zh) 网元管理方法、网管系统、独立计算节点、计算机设备、存储介质
CN106021512A (zh) 一种页面刷新方法及装置
Jiang et al. Neural combinatorial optimization for energy-efficient offloading in mobile edge computing
Liu et al. Hastening stream offloading of inference via multi-exit dnns in mobile edge computing
WO2023185822A1 (zh) 网元注册方法、模型确定方法、装置、网元、通信系统及存储介质
WO2024011376A1 (zh) 人工智能ai网络功能服务的任务调度方法及装置
WO2023185826A1 (zh) 网元注册方法、模型请求方法、装置、网元、通信系统及存储介质
WO2023207980A1 (zh) 模型信息获取方法、发送方法、装置、节点和储存介质
CN114443234A (zh) 数据分析方法、装置、nwdaf组群及可读存储介质
CN111431743B (zh) 基于数据分析的大规模WiFi系统中边缘资源池构建方法及系统
CN116828445A (zh) 网元注册方法、模型确定方法、装置、网元、通信系统及存储介质
Anitha et al. A generic resource augmentation architecture for efficient mobile communication
WO2023169425A1 (zh) 通信网络中的数据处理方法及网络侧设备
RU2465637C2 (ru) Система и способ улучшения работы медиасервера
WO2023213246A1 (zh) 模型选择方法、装置及网络侧设备
WO2023213288A1 (zh) 模型获取方法及通信设备
CN106027344A (zh) 基于大数据的家居服务系统
WO2023185788A1 (zh) 候选成员的确定方法、装置及设备
WO2024008154A1 (zh) 联邦学习方法、装置、通信设备及可读存储介质
WO2024061252A1 (zh) 信息获取方法、装置、网络侧设备及存储介质
WO2024061254A1 (zh) 数据获取方法、装置、系统和设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23778187

Country of ref document: EP

Kind code of ref document: A1