Model information acquisition method, transmission method, device, node, and storage medium

Abstract

This application discloses a model information acquisition method, a transmission method, an apparatus, a node, and a storage medium, and belongs to the technical field of communications. The model information acquisition method according to an embodiment of this application includes: a step in which a model training functional node determines a federated learning FL server-side node; a step in which the model training functional node transmits a first request message for triggering the federated learning of the FL server-side node so as to acquire a target model to the FL server-side node; and a step in which the model training functional node receives information on the target model transmitted by the FL server-side node.

Description

[Technical Field]

[0001] (CROSS-REFERENCE TO RELATED APPLICATIONS) This application claims priority to a Chinese patent application filed in China on April 29, 2022, bearing application number 202210476336.X, the entire contents of which are incorporated herein by reference.

[0002] The present application relates to the technical field of communications, and specifically to a model information acquisition method, transmission method, device, node, and storage medium. [Background technology]

[0003] Currently, some communication systems (e.g., 5th-generation mobile communication (5G)) have introduced artificial intelligence (AI) functions, specifically, models that process network services. However, because only the independent learning method of network nodes is currently supported to acquire models, in some cases (e.g., when training data is insufficient), network nodes may not be able to acquire models through training, resulting in poor model training performance of network nodes. Summary of the Invention

[0004] The embodiments of the present application provide a model information acquisition method, a transmission method, an apparatus, a node, and a storage medium to solve the problem of poor model training performance of network nodes.

[0005] In the first aspect, a model training function node determining a federated learning (FL) server-side node; The model training function node sends a first request message to the FL server side node to trigger federated learning of the FL server side node to obtain a target model; and a step in which the model training function node receives information of the target model sent by the FL server-side node.

[0006] In a second aspect, a step of receiving a first request message sent by a model training function node by a federated learning FL server side node, the first request message being for triggering federated learning of the FL server side node to obtain a target model; The FL server node performs federated learning with the FL client node based on the first request message to obtain the target model; the FL server-side node transmitting information of the target model to the model training function node.

[0007] In a third aspect, a first determination module for determining a federated learning FL server-side node; a first sending module for sending a first request message to the FL server side node to trigger federated learning of the FL server side node to obtain a target model; a receiving module for receiving information about the target model transmitted by the FL server-side node.

[0008] In a fourth aspect, a receiving module for receiving a first request message sent by a model training function node, the first request message being for triggering federated learning of a federated learning FL server side node to obtain a target model; a learning module for performing federated learning with an FL client node based on the first request message to obtain the target model; a transmitting module for transmitting information of the target model to the model training function node.

[0009] In a fifth aspect, there is provided a model training function node including a processor and a memory, wherein a program or command executable by the processor is stored in the memory, and when the program or command is executed by the processor, steps of the model information acquisition method provided in the embodiments of the present application are realized.

[0010] In a sixth aspect, there is provided a model training function node including a processor and a communication interface, wherein the processor or the communication interface is used to determine an FL server side node, and the communication interface is used to send a first request message to the FL server side node to trigger federated learning of the FL server side node to obtain a target model, and to receive information about the target model sent by the FL server side node.

[0011] In a seventh aspect, there is provided a server-side node including a processor and a memory, wherein a program or command executable by the processor is stored in the memory, and when the program or command is executed by the processor, steps of the model information transmission method provided in the embodiments of the present application are realized.

[0012] In an eighth aspect, there is provided a server-side node including a processor and a communication interface, wherein the communication interface is used to receive a first request message sent by a model training function node, the first request message being for triggering federated learning of a federated learning FL server-side node to obtain a target model, the processor or communication interface is used to perform federated learning with an FL client node based on the first request message to obtain the target model, and the communication interface is used to send information of the target model to the model training function node.

[0013] In a ninth aspect, there is provided a model information transmission system including a model training function node and a server-side node, wherein the model training function node can be used to perform steps of a model information acquisition method provided in an embodiment of the present application, and the server-side node can be used to perform steps of a model information transmission method provided in an embodiment of the present application.

[0014] In a tenth aspect, there is provided a readable storage medium having a program or command stored thereon, which, when executed by a processor, realizes steps of a model information acquisition method provided in an embodiment of the present application, or realizes steps of a model information transmission method provided in an embodiment of the present application.

[0015] In an eleventh aspect, there is provided a chip including a processor and a communication interface, wherein the communication interface and the processor are coupled, and the processor is configured to execute a program or command to realize the model information acquisition method provided in the embodiments of the present application, or to realize the model information transmission method provided in the embodiments of the present application.

[0016] In a twelfth aspect, there is provided a computer program / program product that is stored in a storage medium and that, when executed by at least one processor, implements steps of the model information acquisition method provided in the embodiments of the present application or implements the model information transmission method provided in the embodiments of the present application.

[0017] In an embodiment of the present application, a model training function node determines an FL server-side node, and the model training function node sends a first request message to the FL server-side node to trigger federated learning of the FL server-side node to obtain a target model, and the model training function node receives the target model information sent by the FL server-side node. In this way, the model training function node obtains the target model information through federated learning, and can improve the model training performance of the network node. [Brief explanation of the drawings]

[0018] [Figure 1] 1 is a block diagram of a wireless communication system to which an embodiment of the present application can be applied. [Figure 2] 1 is a flowchart of a model information acquisition method provided in an embodiment of the present application; [Figure 3] 1 is a flowchart of a model information transmission method provided in an embodiment of the present application; [Figure 4] FIG. 1 is a schematic diagram of a model information acquisition method provided in an embodiment of the present application. [Figure 5] FIG. 1 is a schematic diagram of another model information acquisition method provided in an embodiment of the present application. [Figure 6] FIG. 1 is a structural diagram of a model information acquisition device provided in an embodiment of the present application; [Figure 7] FIG. 2 is a structural diagram of a model information transmitting device provided in an embodiment of the present application; [Figure 8] FIG. 1 is a structural diagram of a communication device provided in an embodiment of the present application; [Figure 9] FIG. 2 is a structural diagram of a network node provided in an embodiment of the present application; DETAILED DESCRIPTION OF THE INVENTION

[0019] Hereinafter, the technical solutions in the embodiments of the present application will be clearly explained with reference to the drawings in the embodiments of the present application, and it should be understood that the described embodiments are only a part of the embodiments of the present application, not all of the embodiments, and all other embodiments obtained by those skilled in the art based on the embodiments in the present application fall within the scope of protection of the present application.

[0020] The terms "first," "second," etc., used in the specification and claims of this application are not intended to describe a particular order or chronology, but rather to distinguish between similar objects. It should be understood that terms used in this manner may be interchanged where appropriate so that the embodiments of this application can be implemented in orders other than those illustrated or described herein. It should also be understood that the objects distinguished by "first" and "second" generally refer to one type and do not limit the number of objects; for example, the first object may be one or multiple. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the " / " symbol generally indicates that the related objects before and after are in an "or" relationship.

[0021] It should be noted that the techniques described in the embodiments of the present application are not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, but can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-Carrier Frequency Division Multiple Access (SC-FDMA), and other systems. The terms "system" and "network" in the embodiments of the present application are often used interchangeably, and the described techniques can be used in other systems and wireless technologies in addition to those mentioned above. Although the following description describes New Radio (NR) systems for illustrative purposes and uses NR terminology in much of the following description, these techniques may also be applied to applications other than NR system applications, such as 6th Generation (6G) communication systems.

[0022] 1 shows a block diagram of a wireless communication system to which an embodiment of the present application can be applied. The wireless communication system includes a terminal 11 and a network side device 12. The terminal 11 may be a terminal side device such as a mobile phone, a tablet personal computer (TPC), a laptop computer (LC), also known as a notebook computer, a personal digital assistant (PDA), a personal digital assistant, a netbook, an ultra-mobile personal computer (UMPC), a mobile internet device (MID), an augmented reality (AR) / virtual reality (VR) device, a robot, a wearable device, a vehicle user equipment (VUE), a pedestrian user equipment (PUE), a smart home (home devices equipped with wireless communication functions such as a refrigerator, a television, a washing machine, or furniture), a game console, a personal computer (PC), an automated teller machine, or a kiosk. The wearable device includes a smart watch, a smart wristband, a smart earphone, a smart glass, smart jewelry (such as a smart bangle, a smart bracelet, a smart ring, a smart necklace, a smart anklet bangle, a smart anklet, etc.), a smart wrist strap, smart wear, etc. It should be noted that the specific type of the terminal 11 is not limited in the embodiments of the present application.

[0023] The network side equipment 12 may include an access network device and a core network device, and the access network device may be called a radio access network device, a radio access network (RAN), a radio access network function, or a radio access network unit. The access network device may include a base station, a wireless local area network (WLAN) access point, or a wireless fidelity (WiFi) node, and 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, a radio transceiver, a basic service set (BSS), an extended service set (ESS), a home B node, a home evolved B node, a transmitting receiving point (TRP), or any other appropriate term in the field. As long as the same technical effect can be achieved, the base station is not limited to a specific technical term. It should be noted that in the embodiments of this application, only a base station in an NR system is described as an example, and the specific type of the base station is not limited.

[0024] Core network devices include core network nodes, core network functions, Mobility Management Entity (MME), Access and Mobility Management Function (AMF), Session Management Function (SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Server Discovery Function (EASDF), Unified Data Management (UDM), Unified Data Repository (UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (or L-NEF), Binding Support Function (BSF), and Application Function (A-NEF). The core network device may include, but is not limited to, at least one of a core network function (Function (AF)), a network data analytics function (NWDAF), etc. It should be noted that in the embodiments of the present application, only core network devices in an NR system are described as examples, and the specific type of core network device is not limited.

[0025] In the embodiments of the present application, the model training function node may be a network node for generating a model and performing model training, and may refer to a general model training function node (not an FL server-side node) or an FL client node; that is, the fact that the model training function node is not an FL server-side node may be understood to mean that the model training function node cannot be used as an FL server-side node for a specific model training task (corresponding to a specific analysis identifier, a specific Area of ​​Interest (AOI)), but can be used as an FL server-side node for other model training tasks (e.g., other analysis identifiers, other AOIs), but is not limited to this case.

[0026] The model inference function node may be a network node for inferring and generating predictive information, generating statistical information, or performing data analysis.

[0027] In some embodiments, the model training function node may refer to a network element, terminal, or module in a communication network that has an AI model training function, and the model inference function node refers to a network element, terminal, or module in a communication network that has a model inference function. That is, the model training function node and the model inference function node may naturally be called by other names.

[0028] In the embodiments of the present application, the model training function node and the model inference function node may be core network elements or modules within core network elements. For example, the NWDAF may include a model training function node and a model inference function node, the FL server side node may be a core network element with federated learning server functionality, and the FL client node may be another core network element or module participating in federated learning. Alternatively, the model training function node and the model inference function node may be radio access network elements or modules within radio access network elements, i.e., functions within the RAN. For example, the model training function node may be a RAN device with model training functionality, specifically, for example, a base station device or module with model training functionality but that is not an FL server side node. The FL server side node may be a base station device or module or operation, administration, and maintenance (OAM) with federated learning server functionality, and the FL client node may be another base station device or module that is a member participating in federated learning.

[0029] Furthermore, the model training function node and the model inference function node may be terminals or functions within a terminal. For example, the model training function node may be a terminal having a model training function. Specifically, for example, it may be a terminal having a model training function but not an FL server-side node. The FL server-side node may be a terminal or AF having the function of a federated learning server, and other FL client nodes may be other terminals that become members participating in federated learning.

[0030] It should be noted that in the embodiments of the present application, the model training function node and the model inference function node may be independently arranged as different network element devices, or may be jointly arranged as two function modules within the same network element device, such as a core network element, a radio access network element, or a terminal, in which case the core network element, the radio access network element, or the terminal can provide not only the AI ​​model training function, but also the model inference function.

[0031] In the embodiments of the present application, the FL server-side node is a network element that coordinates or manages federated learning, and may be, for example, an FL center network element or an FL coordinator network element, or a model training function central node that can be used in FL operation, and may specifically be a network element such as a core network element, a radio access network element, a terminal, or an application server. The FL client node is a network element that participates in federated learning, and may also be called an FL operation participating network element, and may specifically be a network element such as a core network element, a radio access network element, a terminal, or an application server.

[0032] In some embodiments, the model training function node may be a Model Training Logical Function (MTLF), the model inference function node may be an Analytics Logical Function (AnLF), the FL server-side node may be an FL server, and the FL client node may be an FL client.

[0033] Referring to FIG. 2, which is a flowchart of the model information obtaining method provided in the embodiment of the present application, as shown in FIG. 2, the method includes the following steps 201 to 203.

[0034] In step 201, the model training function node determines the FL server side node.

[0035] In the step of determining the FL server-side node, the model training function node may inquire about the FL server-side node from other network nodes, or the model training function node may select the FL server-side node according to pre-acquired setting information.

[0036] In step 202, the model training function node sends a first request message to the FL server node to trigger federated learning of the FL server node to obtain a target model.

[0037] The first request message may request the FL server node to perform federated learning and training to obtain the target model. Specifically, the FL server may perform FL operations with at least one FL client node, for example, the FL server may perform an interactive iterative process of federated learning with at least one FL client node to obtain the target model. Since the target model is a model obtained through federated learning, it may be referred to as a federated learning model.

[0038] In step 203, the model training function node receives the information of the target model sent by the FL server side node.

[0039] In this step, the FL server node may transmit information of the target model to a model training function node after obtaining the target model through federated learning.

[0040] The information on the target model may be information for determining the target model, such as model file information of the target model or download address information of the model file.

[0041] It should be noted that in the embodiments of the present application, the target model may be a model for processing communication services in a communication system, such as a model for data analysis, or a model for inference tasks, or a model for channel estimation, or a model for information prediction, etc.

[0042] In an embodiment of the present application, the above steps enable the model training function node to obtain information on the target model through federated learning, thereby improving the model training performance of the network node and solving the problem of not being able to obtain the target model due to data privacy reasons. For example, when the model training function node finds that it needs to provide a target model for the AOI specified by the model inference function node, but may not be able to obtain all or part of the training data within the AOI due to a data island problem, it triggers federated learning of the FL server-side node.

[0043] In one alternative embodiment, the step of the model training function node determining the FL server side node comprises: The model training function node sends a node discovery request message to a network repository function network element to request network nodes to participate in federated learning training; The model training function node receives a response message sent by the network repository function network element, the response message including information of the FL server side node.

[0044] Here, the node discovery request message may be a network element discovery request (Nnrf_NFDiscovery_Request).

[0045] The network repository function network element can store information of one or more FL server side nodes, and after receiving the request message, returns the information of the corresponding FL server side node to the model training function node.

[0046] In this embodiment, the network repository function network element can realize information acquisition of the FL server side node.

[0047] It should be noted that in some embodiments, the information of the FL server side node may not be obtained using a network repository function network element; for example, the information of the FL server side node may be fixedly set within the model training function node.

[0048] Optionally, the node discovery request message includes: The information includes at least one of an analytics ID, an area of ​​interest (AOI) information, an interest time information, a model description method information, a model sharable information, a model performance information, a model algorithm information, a model training speed information, a federated learning instruction information, a federated learning type information, an FL server side node type instruction information, an FL client node type instruction information, a first service information, and a second service information.

[0049] The analysis identifier may be used to indicate that the network node requested by the request message should support the model training task corresponding to the analysis identifier.

[0050] The AOI information can be used to indicate that the network node requested by the request message is capable of providing service to an area, which may be at least one Tracking Area (TA), at least one cell, or other area, corresponding to the AOI information.

[0051] The time of interest information can be used to indicate that the network node requested by the request message should support model training for the time period corresponding to the time of interest information.

[0052] The model description method information, which may also be referred to as model description method requirement information or model description method expectation information, can be used to indicate that the network node requested by the request message should support a model representation based on a model description method corresponding to the model description method information. For example, the model description method information may be a model representation language represented by Open Neural Network Exchange (ONNX) or a model framework represented by TensorFlow, Pytorch, or the like.

[0053] The model shareability information, which may also be referred to as model shareability requirement information or model shareability expectation information, can be used to indicate that the network node requested by the request message must be able to share a model with the model training function node, where shareability means being interoperable, or being able to understand and execute each other.

[0054] The model performance information, which may also be referred to as model performance requirement information or model performance information expectation information, can be used to indicate that the network node requested by the request message needs to provide a model that can satisfy the model performance information, where the performance may be a model accuracy value, a mean absolute error (MAE), etc.

[0055] The model algorithm information, which may also be referred to as model algorithm requirement information or model algorithm expectation information, can be used to indicate that the network node requested by the request message should support training of a model based on the model algorithm corresponding to the model algorithm information.

[0056] The model training speed information, which may also be referred to as model training speed requirement information or model training speed expectation information, can be used to indicate that the speed of model training by the network node requested by the request message must satisfy the model training speed indicated by the model training speed information, where the training speed may be expressed as the time it takes for model training to converge or reach a certain performance threshold.

[0057] The federated learning indication information can be used to indicate that the network node requested by the request message should support federated learning.

[0058] The federated learning type information indicates that the type of federated learning that the network node requested by the request message should support is: It can be used to indicate at least one of horizontal associative learning and vertical associative learning.

[0059] Here, the type of horizontal associative learning may be to perform learning training using different training data samples with the same feature points, and the type of vertical associative learning may be to perform learning training using training data samples with different feature points of the same training sample.

[0060] The FL server side node type indication information can be used to indicate that the network node requested by the request message belongs to the FL server side node type.

[0061] The FL client node type indication information can be used to indicate that the network node requested by the request message belongs to an FL client node type.

[0062] The first service information can be used to indicate that the network node requested by the request message should support the service of the federated learning server.

[0063] The second service information can be used to indicate that the network node requested by the request message should support the service of the federated learning member.

[0064] In this embodiment, since the node discovery request message includes the above-mentioned at least one information, the network repository function network element can return information of the FL server side node that meets the corresponding condition to the model training function node, so that the FL server side node can finally obtain the required, acceptable or desired target model of the above-mentioned model training function node through federated learning.

[0065] Optionally, the response message includes information of N network nodes including the FL server side node, where N is a positive integer; The information for each network node is The information includes at least one of a fully qualified domain name (FQDN), identifier information, and address information.

[0066] In this embodiment, the FL server side node is determined by at least one of the FQDN, identifier information, and address information.

[0067] In some embodiments, the N network nodes further include an FL client node.

[0068] One or more FL client nodes included in the N network nodes may be FL client nodes participating in this federated learning, and these FL client nodes may be FL participants or FL members.

[0069] Since the response message further includes the FL client node, the model training function node can quickly determine the FL server node and the FL client node, thereby improving the efficiency of obtaining the target model by timely requesting the FL server node and the FL client node to perform federated learning.

[0070] Optionally, the information for each network node is: The network node further includes type information for indicating a type of the network node, the type being one of an FL server node and an FL client node.

[0071] In this embodiment, the type information can indicate that the network node is an FL server node or an FL client node. Of course, in the embodiment of the present application, the type information does not need to be indicated. For example, the model training function node can identify whether it is an FL server node or an FL client node according to the identifier information of these network nodes. Or, if the response message includes the FL server node and the FL client node, the information of the FL server node and the FL client node can be sorted in the response message, so that the model training function node can identify the FL server node and the FL client node by sorting the information.

[0072] In one alternative embodiment, the first request message comprises: including at least one of associative learning instruction information and a model identifier; the federated learning instruction information is for requesting the FL server side node to trigger federated learning to obtain a target model; The model identifier is for uniquely identifying the target model.

[0073] Here, the model identifier may be obtained by the model training function node. For example, the method further includes the step of the model training function node obtaining the model identifier, for example, by self-generating the model identifier or obtaining it from a model identifier management network element.

[0074] In this embodiment, the first request message includes at least one of federated learning instruction information and a model identifier, so that the FL server node can quickly respond to the request message by deciding to perform federated learning. It should be noted that in some embodiments, the federated learning instruction information and the model identifier may not be carried, for example, the model identifier may be obtained by the FL server node, so that the FL server node will perform federated learning by default when receiving the first request message.

[0075] In some embodiments, the first request message comprises: and may include at least one of an analysis identifier, time of interest information, model description method information, model shareability information, model performance information, model algorithm information, model training speed information, first model filter information, first target information, and report information; For the above analysis identifier, interest time information, model description method information, model shareability information, model performance information, model algorithm information, and model training speed information, please refer to the corresponding descriptions in the above embodiments, and detailed descriptions will be omitted here.

[0076] The first model filter information can be used to indicate at least one of a domain range, a time range, a slice range, and a data network name (DNN) range for the associative learning.

[0077] The first target information can be used to indicate that the federated learning targets one of a single terminal, multiple terminals, or any terminal, and for example, the first target information includes terminal information such as a terminal identifier (UE ID) or a terminal group identifier (UE group ID).

[0078] The reporting information can be used to direct the reporting format and / or reporting conditions of the federation model transmission.

[0079] In this embodiment, the first request message includes the at least one piece of information, so that the FL server side node can finally obtain the required, acceptable or desired target model of the model training function node through federated learning.

[0080] In one alternative embodiment, the first request message includes information of FL client nodes that will participate in federated learning.

[0081] The FL client nodes participating in the federated learning may be one or more FL client nodes, and these clients may or may not include the model training function node.

[0082] The information of the FL client nodes used in the federated learning may be obtained from a network repository function network element or may be preset by a model training function node. Since the first request message includes information of the FL client nodes participating in the federated learning, the FL server node can quickly perform federated learning with the FL client nodes, thereby improving the efficiency of acquiring the target model.

[0083] In an alternative embodiment, the target model information is: The target model includes at least one of a model identifier, association instruction information, a model file, and address information of the model file.

[0084] The associative learning instruction information is used to indicate that the target model is a model obtained by associative learning.

[0085] The model identifier is for uniquely identifying the target model.

[0086] The model file may include information about the target model's associated files, such as files for the model's network structure, weight parameters, and input / output data.

[0087] The address information of the model file is address information for acquiring the target model, such as a storage address for indicating the model file, or download address information of the model file.

[0088] It should be noted that in the embodiments of the present application, the target model information does not need to include the model identifier, since the model training function node can obtain the identifier by itself; it does not need to include the federated instruction information, since the model training function node can recognize by default that the target model sent by the FL server-side node is a model obtained by federated learning; it does not need to include the model file, since the model training function node downloads the model file based on the address information; and in some embodiments, the model file can be stored in a preset location after the model is trained, so the model training function node can download the model file from this location, so it does not need to include the address information of the model file.

[0089] In some embodiments, the target model information comprises: The information may further include at least one of an analysis identifier, second model filter information, and model validity information corresponding to the target model.

[0090] The analysis identifier can be used to indicate the task corresponding to the target model.

[0091] The second model filter information can be used to indicate at least one of a domain range, a time range, a slice range, and a Data Network Name (DNN) range of the target model.

[0092] The model validity information can be used to indicate validity information such as validity time and validity area of ​​the model.

[0093] In an alternative embodiment, the method comprises: The model training function node further includes sending information about the target model to a model inference function node.

[0094] In this embodiment, the information of the target model may be sent to the model inference function node spontaneously, or may be sent based on the request of the model inference function node.

[0095] For example, in a request transmission mode, the model training function node receives a model request message transmitted by a model inference function node, the model request message comprising: and may include at least one of an analysis identifier, third model filter information, second target information, and time information; The analysis identifier can be used to indicate the data analysis task targeted by the model identifier requested by the model request message.

[0096] The third model filter information can be used to indicate conditions that the model requested by the model request message must satisfy, for example, the conditions may be a region of interest that the model needs to serve, Single Network Slice Selection Assistance Information (S-NSSAI), or a DNN.

[0097] The second target information can be used to indicate a training target of the model requested by the model request message, and the training target may be a single terminal, multiple terminals, any terminal, etc. For example, the second target information may be a terminal identifier or a terminal group identifier of the terminal.

[0098] The time information is for indicating at least one of an application time of the model requested by the model request message and a reporting time of the model information.

[0099] In this embodiment, the model training function node sends information of the target model to the model inference function node, so that the model inference function node can use the target model to improve service performance.

[0100] In one alternative embodiment, before the step of the model training function node determining the FL server side node, the method further comprises: The method further includes the step of the model training function node determining that the target model should be obtained by federated learning.

[0101] The step of determining that the target model should be obtained by federated learning may be performed by a model training function node determining that the target model should be obtained by federated learning because the target model cannot be obtained by independent training. For example, the step of determining that the target model should be obtained by federated learning may be performed by a model training function node determining that the target model should be obtained by federated learning. If the model training function node determines that all or part of the training data for generating the target model cannot be obtained, the method includes determining that the target model should be obtained by federated learning.

[0102] The inability to obtain all or part of the training data for generating the target model may be due to reasons such as technical confidentiality or user privacy. For example, when the model training function node needs to provide a target model for an AOI specified by the model inference function node but finds that it may be unable to obtain all or part of the training data within the AOI due to a data island issue, it triggers federated learning at the FL server-side node. For example, when the model training function node needs to obtain models corresponding to some terminals but is unable to obtain training data corresponding to these terminals due to data privacy issues, it may rely on the FL server-side node to obtain training data corresponding to these terminals and trigger federated learning to obtain a target model for the terminal.

[0103] In this embodiment, if it is determined that the target model should be obtained by federated learning, the FL server node can obtain information about the target model, thereby improving the model training performance of the network node. For example, the model training function node can obtain the target model even if it does not have enough training data for the target model, and can obtain the target model even if it does not have enough training resources for the target model, thereby improving the model training performance of the network node.

[0104] In an embodiment of the present application, the model training function node determines an FL server-side node, and the model training function node sends a first request message to the FL server-side node to trigger federated learning of the FL server-side node to obtain a target model, and the model training function node receives the target model information sent by the FL server-side node. In this way, the model training function node can obtain the target model information through federated learning, thereby improving the model training performance of the network node.

[0105] Referring to FIG. 3, which is a flowchart of the model information transmission method provided in the embodiment of the application, as shown in FIG. 3, the method includes the following steps 301 to 303.

[0106] In step 301, an FL server-side node receives a first request message sent by a model training function node, the first request message being for triggering federated learning of the FL server-side node to obtain a target model.

[0107] In step 302, the FL server node performs federated learning with the FL client node based on the first request message to obtain the target model.

[0108] Here, the step of the FL server node performing federated learning with the FL client node based on the first request message may trigger the execution of an FL operation between the FL server node and the FL client node by the first request message; specifically, the FL server node and the FL client node may train and obtain the target model by performing an interactive iterative process of federated learning.

[0109] In step 303, the FL server node sends information about the target model to the model training function node.

[0110] Optionally, the first request message comprises: including at least one of associative learning instruction information and a model identifier; the federated learning instruction information is for requesting the FL server side node to trigger federated learning to obtain a target model; The model identifier is for uniquely identifying the target model.

[0111] In some embodiments, the first request message comprises: The information may include at least one of an analysis identifier, time of interest information, model description method information, model shareability information, model performance information, model algorithm information, model training speed information, first model filter information, first target information, and report information.

[0112] Optionally, the first request message includes information of FL client nodes that will participate in federated learning.

[0113] Optionally, the method further comprises: The method further includes a step in which the FL server-side node determines FL client nodes that will participate in federated learning.

[0114] Here, in the above step in which the FL server side node determines the FL client nodes to participate in federated learning, the FL server side node may query the FL client nodes to participate in federated learning from a network repository function network element, or the FL server side node may determine the FL client nodes to participate in federated learning based on pre-configured information.

[0115] Optionally, the step of the FL server side node determining the FL client nodes to participate in the federated learning includes: The FL server side node sends a node discovery request message to a network repository function network element to request FL client nodes to participate in federated learning; The FL server side node includes a step of receiving a response message sent by the network repository function network element, the response message including information of the FL client nodes participating in federated learning.

[0116] Optionally, the federated model information comprises: At least one of a model identifier, association instruction information, a model file, and address information of the model file corresponding to the target model is included; the associative learning instruction information is for instructing that the target model is a model acquired by associative learning, The model identifier is for uniquely identifying the target model.

[0117] In some embodiments, the target model information comprises: The information may further include at least one of an analysis identifier, second model filter information, and model validity information corresponding to the target model.

[0118] Optionally, the model identifier is the one obtained by the model training function node for the target model.

[0119] Optionally, the method further comprises: The FL server-side node further includes obtaining the model identifier for the target model.

[0120] Here, the model identifier obtained by the FL server node for the target model may be generated by the FL server node or obtained from a model identifier management network element.

[0121] It should be noted that this embodiment is an embodiment of an FL server side node corresponding to the embodiment shown in Fig. 2, and for a specific embodiment, please refer to the related description of the embodiment shown in Fig. 2. In order to avoid repetition, detailed description will be omitted in this embodiment.

[0122] In the following, the model training function node is MTLF, the model inference function node is AnLF, the FL server side node is FL server, and the FL client node is FL client as an example, and the methods provided in the embodiments of the present application will be described through several embodiments.

[0123] Example 1 In this embodiment, the determination of an FL client by an FL server will be described as an example. Specifically, as shown in Figure 4, the process includes the following steps 0 to 2, steps 3a to 3b, steps 4 to 7, step 7a, and steps 8 to 11.

[0124] In step 0, optionally, a consumer network function (Consumer NF) requests data analysis results from an AnLF, and the request message can carry at least one of an analysis identifier, analysis filter information, analysis report target, and analysis report information.

[0125] The analytics identifier (Analytics ID) is used to identify the type of data analysis task. For example, analytics ID=UE mobility indicates a task that requires analysis of user movement trajectory data.

[0126] Analytic filter information is for indicating data analysis result filter information including area of ​​interest (AOI), slice S-NSSAI, DNN, etc.

[0127] The target of analytic reporting is used to specify whether the target of data analysis is a terminal, multiple terminals, or all UEs.

[0128] The analytics reporting information is used to indicate information such as the reporting time and reporting conditions for the data analysis results.

[0129] In step 1, the AnLF sends a model request message to the MTLF1 to request a model corresponding to the analysis identifier.

[0130] The model request message may be a model provision subscription (Nnwdaf_MLModelProvision_subscribe) message or a model information request message (Nnwdaf_MLModelInfo_Request).

[0131] The model request message may carry an Analytic ID, model filter information, model target information, and model time information.

[0132] The analytics ID is used to identify the type of inference task. For example, analytics ID=UE mobility is used to predict a user movement trajectory.

[0133] The model filter information is to indicate the conditions that the required model should satisfy, such as area of ​​interest (AOI), slice S-NSSAI, DNN, etc.

[0134] The model target information is for indicating a model training target, such as a single terminal, multiple terminals, or any terminal.

[0135] The model time information is for indicating the application time of the model or the reporting time of the model information.

[0136] It should be noted that the MTLF1 here is not an FL server, specifically not an FL server of the target model, but may be, but is not limited to, an FL server of another model, and the FL server may be called a coordinator or a central MTLF for FL operation, but is merely one FL client (or called a member participating in FL operation) or an MTLF that does not support FL capabilities.

[0137] In step 2, MTLF1 determines, based on the model request message, that it needs to perform federated learning to obtain the requested model.

[0138] Factors that may cause MTLF1 to determine that associative learning needs to occur include:

[0139] In MTLF1, there is insufficient training data for the model training task, e.g., due to data privacy concerns, specific training data is lacking in other domains or other training entities, and intensive model training cannot be performed to generate the required model.

[0140] Alternatively, the MTLF1 may determine that the type of associative learning is horizontal associative learning. For example, the MTLF1 may determine that horizontal associative learning is required based on the fact that the training task satisfies the characteristic that the samples corresponding to the training data are different but have the same features.

[0141] In step 3a, optionally, the MTLF1 sends a node discovery request message to the NRF to request network element devices capable of performing federated learning and training.

[0142] Specifically, the node discovery request message can be a discovery request (Nnrf_NFDiscovery_Request), which may include an analysis identifier, AOI.

[0143] The analytics ID is used to indicate that the requested network element should support the analytics ID.

[0144] The AOI may be at least one Tracking Area (TA), at least one cell, or other representation to indicate that the required network element must be able to provide service to the AOI.

[0145] The node discovery request message may further include at least one of federated learning instruction information, information that the expected NF type is an FL server, information that the expected NF type is an FL client, information that the expected NF service name is a collaborative service FL service, and information that the expected NF service name is a service joining FL client.

[0146] The federated learning indication information is for indicating that the required network element must be able to support federated learning, and can also be used to indicate that it should support horizontal federated learning.

[0147] The information that the expected NF type is an FL server (Expected NF type=FL server) is to indicate that the requested network element should belong to the FL server type.

[0148] The information that the expected NF type is an FL client (Expected NF type=FL client) is to indicate that the requested network element should belong to the FL client type.

[0149] The information that the expected NF service name is a coordination service FL service (Expected NF service Name=FL server (coordination) service) is to indicate that the requested network element should support the federated learning server (coordination) service.

[0150] The information Expected NF service Name = FL client (participant) service is to indicate that the requested network element should support the service of the federation learning member (participant).

[0151] In one implementation, the network element request message, which can specify an analysis identifier, an expected NF type=MTLF, and an AOI, includes federated learning instruction information, in this case MTLF1, to request from the NRF all MTLFs that support federated learning for the AOI.

[0152] In one implementation, the network element request message includes information that the expected NF type is an FL server (Expected NF type=FL server), and is otherwise similar to related technologies (e.g., specifying an analytics ID, Expected NF type=MTLF, and AOI), where MTLF1 is used to request an FL server that supports federated learning from the NRF.

[0153] In one implementation, the network element request message includes information such as Expected NF type=FL server and FL client, and other information is similar to that of the related technology (e.g., analytics ID, Expected NF type=MTLF, AOI specification), where MTLF1 is for requesting an FL server and FL client that support federated learning from the NRF.

[0154] In step 3b, the NRF returns to the MTLF1 the equipment information that satisfies the request message initiated by the network element in step 3a.

[0155] The NRF sends the FQDN, identifier information, address information, etc. of one or more MTLFs that meet the request to MTLF1, and the feedback information can specify for each MTLF whether it is an FL server or an FL client.

[0156] In step 4, optionally, MTLF1 assigns a model identifier (model ID) to the associative learning model to be generated in order to uniquely identify the model.

[0157] In step 5, the MTLF1 sends a federated learning request message to the FL server to request the FL server to trigger a federated learning process, where the federated learning request message may include at least one of the following information: a federated learning instruction, an analysis identifier, a model identifier, model filter information, a model target, and model report information.

[0158] The associative learning indication (FL indication) is for indicating a request to perform an associative learning process.

[0159] The analytics ID is used to request a task type for the analytics ID identifier and to instruct the federated learning process.

[0160] The model identifier (Model ID) is used to uniquely identify a model created through federated learning.

[0161] Model filter information is used to limit the scope of the federated learning process, such as domain range, time range, S-NSSAI, DNN, etc.

[0162] The model target is used to specify the target of the federated learning process, such as one or more specific terminals, or all terminals.

[0163] The model reporting information is used to indicate reporting information of the generated federated learning model information, such as reporting time (start time, end time, etc.) and reporting conditions (periodic trigger, event trigger, etc.).

[0164] In step 6, optionally, if a model identifier has not been received from MTLF1, the FL server may assign a model identifier to the federated learning model to be generated to uniquely identify the model.

[0165] In step 7a, the FL server determines at least one FL client to perform this federated learning process. Specifically, the FL server can query the network repository function network element to obtain the FL client to satisfy this federated learning process. See step 3a.

[0166] It should be noted that the FL client here may or may not include MTLF1 itself.

[0167] In step 7, the FL server and the FL client perform an interactive iterative process of federated learning to obtain a federated learning model.

[0168] Here, if the FL client does not include MTLF1 itself, MTLF1 does not participate in the interactive process here.

[0169] In step 8, the FL server sends the target model information of the generated federated learning model to the MTLF1, where the target model information is: A model identifier; Associative learning instructions and The model file (including the model's network structure, weight parameters, input / output data, etc.) Download address information or storage address information of the model file (for indicating the storage address of the model file or the download address of the model file), an analysis identifier (indicating that the model applies to a particular type of inference task); Model filter information (domain range, time range, S-NSSAI, DNN, etc., for limiting the scope of the federated learning process) and Validity area information (area where the model applies), Effective time information (the time when the model applies) and It includes at least one of the following:

[0170] In step 9, MTLF1 sends the model information of the generated federated learning model to AnLF.

[0171] In this step, the MTLF1 can send the model by a model provision notification (Nnwdaf_MLModelProvision_Notify) or a model information response (Nnwdaf_MLModelInfo_Response) message, and the sending content specifically refers to step 8.

[0172] In step 10, AnLF generates data analysis results based on the model.

[0173] In step 11, the AnLF optionally transmits the data analysis results to the consumer NF.

[0174] Example 2 In this embodiment, an example in which the MTLF1 determines the FL client is described, and specifically, as shown in FIG. 5, there are the following differences from the first embodiment.

[0175] In step 3, MTLF1 obtains the FL server and FL client from the NRF.

[0176] In step 5, optionally, the MTLF1 simultaneously notifies the FL server of the FL clients that will participate in this federated learning. The step in which the FL server queries the NRF for the FL clients may be omitted.

[0177] In an embodiment of the present application, if a model needs to be generated by federated learning due to data privacy issues, even if the model training function node itself does not support federated learning capability or federated learning server function, or does not support federated learning or federated learning server function for a specific analysis identifier or a specific AOI, the model training function node can trigger other devices to perform federated learning and obtain the model it needs, thereby expanding the application scope of federated learning and solving data privacy challenges more widely.

[0178] Referring to FIG. 6, which is a structural diagram of a model information acquisition device provided in an embodiment of the present application, as shown in FIG. 6, the model information acquisition device 600 includes: a first determination module 601 for determining a federated learning FL server side node; a first sending module 602 for sending a first request message to the FL server node to trigger federated learning of the FL server node to obtain a target model; and a receiving module 603 for receiving the information of the target model sent by the FL server side node.

[0179] Optionally, the first determination module 601 is used to send a node discovery request message to a network repository function network element to request network nodes to participate in federated learning training, and receive a response message sent by the network repository function network element, wherein the response message includes information of the FL server side node.

[0180] Optionally, the node discovery request message comprises: At least one of an analysis identifier, area of ​​interest (AOI) information, time of interest information, model description method information, model sharability information, model performance information, model algorithm information, model training speed information, federated learning instruction information, federated learning type information, FL server side node type instruction information, FL client node type instruction information, first service information, and second service information is included; the federated learning indication information is for indicating that the network node requested by the request message should support federated learning; the first service information is for indicating that the network node requested by the request message should support a service of a federated learning server; The second service information is for indicating that the network node requested by the request message should support the service of the federated learning member.

[0181] Optionally, the federated learning type information may include information indicating that the type of federated learning that the network node requested by the request message should support is: Horizontal associative learning and Vertical association learning and at least one of.

[0182] Optionally, the response message includes information of N network nodes including the FL server side node, where N is a positive integer.

[0183] The information for each network node is The information includes at least one of a fully qualified domain name (FQDN), identifier information, and address information.

[0184] Optionally, the N network nodes further include an FL client node.

[0185] Optionally, the information for each network node is: The network node further includes type information for indicating a type of the network node, the type being one of an FL server node and an FL client node.

[0186] Optionally, the first request message comprises: including at least one of associative learning instruction information and a model identifier; the federated learning instruction information is for requesting the FL server side node to trigger federated learning to obtain a target model; The model identifier is for uniquely identifying the target model.

[0187] Optionally, the model information acquisition device 600 It further includes an acquisition module for acquiring the model identifier.

[0188] Optionally, the first request message includes information of FL client nodes that will participate in federated learning.

[0189] Optionally, the target model information comprises: At least one of a model identifier, association instruction information, a model file, and address information of the model file corresponding to the target model is included; the associative learning instruction information is for instructing that the target model is a model acquired by associative learning, The model identifier is for uniquely identifying the target model.

[0190] Optionally, the model information acquisition device 600 The method further includes a second sending module for sending information of the target model to a model inference function node.

[0191] Optionally, the model information acquisition device 600 It further includes a second determining module for determining that the target model should be obtained by associative learning.

[0192] Optionally, the second determination module is used to determine that the target model should be obtained by federated learning when the model training function node determines that all or part of the training data for generating the target model cannot be obtained.

[0193] The model information acquisition device can improve the model training performance of a network node.

[0194] The model information acquisition device in the embodiments of the present application may be an electronic device, such as an electronic device equipped with an operating system, or a component of an electronic device, such as an integrated circuit or chip. The electronic device may be a core network device, a network side device, a terminal, or other device other than a terminal.

[0195] The model information acquisition device provided in the embodiments of the present application can implement each process implemented in the method embodiment shown in Figure 2 and achieve the same technical effects. In order to avoid repetition, detailed descriptions will be omitted here.

[0196] Referring to FIG. 7, which is a structural diagram of a model information transmitting device provided in an embodiment of the present application, as shown in FIG. 7, the model information transmitting device 700 includes: a receiving module 701 for receiving a first request message sent by a model training function node, the first request message being for triggering federated learning of a federated learning FL server side node to obtain a target model; a learning module 702 for performing federated learning with an FL client node based on the first request message to obtain the target model; a sending module 703 for sending information of the target model to the model training function node.

[0197] Optionally, the first request message comprises: including at least one of associative learning instruction information and a model identifier; the federated learning instruction information is for requesting the FL server side node to trigger federated learning to obtain a target model; The model identifier is for uniquely identifying the target model.

[0198] Optionally, the first request message includes information of FL client nodes that will participate in federated learning.

[0199] Optionally, the model information transmitting device 700 The network further includes a determination module for determining which FL client nodes will participate in the federated learning.

[0200] Optionally, the determination module is used to send a node discovery request message to a network repository function network element to request FL client nodes to participate in federated learning, and to receive a response message sent by the network repository function network element, wherein the response message includes information of the FL client nodes to participate in federated learning.

[0201] Optionally, the federated model information comprises: At least one of a model identifier, association instruction information, a model file, and address information of the model file corresponding to the target model is included; the associative learning instruction information is for instructing that the target model is a model acquired by associative learning, The model identifier is for uniquely identifying the target model.

[0202] Optionally, the model identifier is the one obtained by the model training function node for the target model.

[0203] Optionally, the model information transmitting device 700 The method further includes an acquisition module for acquiring the model identifier for the target model.

[0204] The model information transmitting device can improve the model training performance of a network node.

[0205] The model information transmission device in the embodiments of the present application may be an electronic device, such as an electronic device with an operating system, or a component of an electronic device, such as an integrated circuit or chip. The electronic device may be a core network device, a network side device, a terminal, or other device other than a terminal.

[0206] The model information transmission device provided in the embodiments of the present application can implement each process implemented in the method embodiment shown in Figure 3 and achieve the same technical effects, so detailed descriptions will be omitted here to avoid repetition.

[0207] Optionally, as shown in Fig. 8, an embodiment of the present application further provides a communication device 800. The communication device 800 includes a processor 801 and a memory 802, and the memory 802 stores programs or commands executable by the processor 801. For example, when the communication device 800 is a first control network element, when the program or command is executed by the processor 801, the steps of the above-mentioned model information acquisition method or the above-mentioned model information transmission method are realized, and the same technical effects can be achieved. In order to avoid repetition, detailed descriptions will be omitted here.

[0208] An embodiment of the present application further provides a model training function node including a processor and a communication interface, wherein the processor or the communication interface is used for determining an FL server side node, and the communication interface is used for sending a first request message to the FL server side node to trigger federated learning of the FL server side node to obtain a target model, and for receiving information of the target model sent by the FL server side node.

[0209] An embodiment of the present application further provides a server-side node including a processor and a communication interface, wherein the communication interface is used to receive a first request message sent by a model training function node, the first request message being for triggering federated learning of a federated learning FL server-side node to obtain a target model, the processor or communication interface being used to perform federated learning with an FL client node based on the first request message to obtain the target model, and the communication interface being used to send information of the target model to the model training function node.

[0210] Specifically, an embodiment of the present application further provides a network node 900. As shown in Fig. 9, the network node 900 includes a processor 901, a network interface 902, and a memory 903. Here, the network interface 902 is, for example, a common public radio interface (CPRI).

[0211] Specifically, the network node 900 of the embodiment of the present application further includes a command or program stored in the memory 903 and executable by the processor 901, and the processor 901 invokes the command or program in the memory 903 to execute the method performed by each module shown in Fig. 6 or Fig. 7, thereby achieving the same technical effect. In order to avoid repetition, detailed description will be omitted here.

[0212] In an embodiment in which the network node is a model training function node, The processor 901 or the network interface 902 is used to determine the federated learning FL server side node; The network interface 902 is used for sending a first request message to the FL server side node to trigger federated learning of the FL server side node to obtain a target model, and for receiving information about the target model sent by the FL server side node.

[0213] Optionally, determining the FL server side node includes: sending a node discovery request message to a network repository function network element to request network nodes to participate in the federated learning training; receiving a response message sent by the network repository function network element, where the response message includes information of the FL server side node.

[0214] Optionally, the node discovery request message comprises: At least one of an analysis identifier, area of ​​interest (AOI) information, time of interest information, model description method information, model sharability information, model performance information, model algorithm information, model training speed information, federated learning instruction information, federated learning type information, FL server side node type instruction information, FL client node type instruction information, first service information, and second service information is included; the federated learning indication information is for indicating that the network node requested by the request message should support federated learning; the first service information is for indicating that the network node requested by the request message should support a service of a federated learning server; The second service information is for indicating that the network node requested by the request message should support the service of the federated learning member.

[0215] Optionally, the federated learning type information may include information indicating that the type of federated learning that the network node requested by the request message should support is: Horizontal associative learning and Vertical association learning and at least one of.

[0216] Optionally, the response message includes information of N network nodes including the FL server side node, where N is a positive integer.

[0217] The information for each network node is The information includes at least one of a fully qualified domain name (FQDN), identifier information, and address information.

[0218] Optionally, the N network nodes further include an FL client node.

[0219] Optionally, the information for each network node is: The network node further includes type information for indicating a type of the network node, the type being one of an FL server node and an FL client node.

[0220] Optionally, the first request message comprises: including at least one of associative learning instruction information and a model identifier; the federated learning instruction information is for requesting the FL server side node to trigger federated learning to obtain a target model; The model identifier is for uniquely identifying the target model.

[0221] Optionally, the processor 901 or the network interface 902 further comprises: Used to obtain the model identifier.

[0222] Optionally, the first request message includes information of FL client nodes that will participate in federated learning.

[0223] Optionally, the target model information comprises: At least one of a model identifier, association instruction information, a model file, and address information of the model file corresponding to the target model is included; the associative learning instruction information is for instructing that the target model is a model acquired by associative learning, The model identifier is for uniquely identifying the target model.

[0224] Optionally, the network interface 902 further comprises: It is used by the model training function node to send information about the target model to the model inference function node.

[0225] Optionally, before the model training function node determines the FL server side node, the processor 901 further It is used to determine that the target model should be obtained by associative learning.

[0226] Optionally, determining that the target model should be obtained by associative learning includes: When the model training function node determines that all or part of the training data for generating the target model cannot be obtained, determining that the target model should be obtained by federated learning.

[0227] In an embodiment where the network node is an FL server node, The network interface 902 is used for receiving a first request message sent by a model training function node, where the first request message is for triggering federated learning of the FL server side node to obtain a target model; performing federated learning with an FL client node based on the first request message to obtain the target model; and sending information of the target model to the model training function node.

[0228] Optionally, the first request message comprises: including at least one of associative learning instruction information and a model identifier; the federated learning instruction information is for requesting the FL server side node to trigger federated learning to obtain a target model; The model identifier is for uniquely identifying the target model.

[0229] Optionally, the first request message includes information of FL client nodes that will participate in federated learning.

[0230] Optionally, the processor 901 or the network interface 902 further comprises: It is used to determine the FL client nodes that will participate in federated learning.

[0231] Optionally, determining FL client nodes to participate in federated learning includes: sending a node discovery request message to a network repository function network element to request FL client nodes to participate in federated learning; receiving a response message sent by the network repository function network element, the response message including information of FL client nodes participating in federated learning.

[0232] Optionally, the federated model information comprises: At least one of a model identifier, association instruction information, a model file, and address information of the model file corresponding to the target model is included; the associative learning instruction information is for instructing that the target model is a model acquired by associative learning, The model identifier is for uniquely identifying the target model.

[0233] Optionally, the model identifier is the one obtained by the model training function node for the target model.

[0234] Optionally, the processor 901 further It is used to obtain the model identifier for the target model.

[0235] It should be noted that in this embodiment, the model training function node and the FL server side node are core network elements as an example.

[0236] An embodiment of the present application further provides a readable storage medium having a program or command stored thereon, which, when executed by a processor, realizes steps of the model information acquisition method provided in the embodiment of the present application, or realizes steps of the model information transmission method provided in the embodiment of the present application.

[0237] The processor is the processor in the terminal described in the above embodiment. The readable storage medium includes a computer readable storage medium such as a computer read only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0238] The embodiments of the present application further provide a chip, which includes a processor and a communication interface, and the communication interface is coupled to the processor, and the processor is used to execute programs or commands to implement the processes of the embodiments of the model information acquisition method or model information transmission method, and can achieve the same technical effects. In order to avoid repetition, detailed descriptions are omitted here.

[0239] It should be understood that the chips referred to in the embodiments of this application may also be referred to as system level chips, system chips, chip systems, or system-on-chips, etc.

[0240] The embodiments of the present application further provide a computer program / program product, which is stored in a storage medium and can be executed by at least one processor to implement the processes of the above-mentioned model information obtaining method or model information transmitting method embodiments and achieve the same technical effects. In order to avoid repetition, detailed descriptions are omitted here.

[0241] An embodiment of the present application further provides a model information transmission system, which includes a model training function node and a server-side node, wherein the model training function node can be used to perform steps of the model information acquisition method provided in the embodiment of the present application, and the server-side node can be used to perform steps of the model information transmission method provided in the embodiment of the present application.

[0242] It should be noted that, as used herein, the terms "comprise," "consist," or any other variation thereof, are intended to include a non-exclusive inclusion, whereby a process, method, article, or apparatus comprising a set of elements includes not only those elements but also other elements not expressly specified or inherent in such process, method, article, or apparatus. Unless otherwise specified, elements qualified by the phrase "comprise..." do not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element. It should also be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may include performing functions substantially simultaneously or in the reverse order, depending on such functionality. For example, the described method may be performed in a different order than described, and various steps may be added, omitted, or combined. Furthermore, features described with reference to one example may be combined in other examples.

[0243] From the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be realized in the form of a combination of software and a necessary common hardware platform, and of course, they can also be realized by hardware, but in many cases the former is a more preferred embodiment. Based on this view, the technical solutions of the present application can be substantially embodied in the form of a computer software product, which is stored in a storage medium (e.g., ROM / RAM, magnetic disk, optical disk) and includes a plurality of commands that cause a terminal (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0244] Although the examples of the present application have been described above with reference to the drawings, the present application is not limited to the above-mentioned specific embodiments, which are merely illustrative and not limiting. Based on the suggestions of the present application, many forms that a person skilled in the art can make without departing from the spirit of the present application and the scope of protection of the claims are all within the scope of protection of the present application.

Claims

1. A model training function node determines a federated learning FL server-side node; The model training function node sends a first request message to the FL server side node to trigger federated learning of the FL server side node to obtain a target model; The model training function node receives information of a target model sent by the FL server side node; The step of the model training function node determining the FL server side node comprises: The model training function node sends a node discovery request message to a network repository function network element to request network nodes to participate in federated learning training; The model training function node receives a response message sent by the network repository function network element, the response message including information of the FL server side node; The node discovery request message A model information acquisition method including at least one of interest time information and FL server side node type indication information.

2. The node discovery request message At least one of an analysis identifier, area of ​​interest (AOI) information, model description method information, model sharable information, model performance information, model algorithm information, model training speed information, federated learning instruction information, federated learning type information, FL client node type instruction information, first service information, and second service information is included; the federated learning indication information is for indicating that the network node requested by the request message should support federated learning; the first service information is for indicating that the network node requested by the request message should support a service of a federated learning server; The method of claim 1 , wherein the second service information is for indicating that the network node requested by the request message should support the service of a federated learning member.

3. The federated learning type information indicates the type of federated learning that the network node requested by the request message should support, Horizontal associative learning and 3. The method of claim 2, wherein the method is for indicating at least one of vertical associative learning.

4. The response message includes information of N network nodes including the FL server side node, where N is a positive integer; The information for each network node is The method of claim 1 , comprising at least one of a fully qualified domain name FQDN, identifier information, and address information.

5. The method of claim 4 , wherein the N network nodes further include an FL client node.

6. The information for each network node is The method of claim 4 or claim 5, further comprising type information for indicating a type of the network node, the type being one of an FL server-side node and an FL client node.

7. The first request message including at least one of associative learning instruction information and a model identifier; the federated learning instruction information is for requesting the FL server side node to trigger federated learning to obtain a target model; the model identifier is for uniquely identifying the target model; The method of claim 1 , further comprising the step of the model training function node obtaining the model identifier.

8. The method according to claim 1 , wherein the first request message includes information of FL client nodes that will participate in federated learning.

9. The information of the target model is: At least one of a model identifier, associative learning instruction information, a model file, and address information of the model file corresponding to the target model is included; the associative learning instruction information is for instructing that the target model is a model acquired by associative learning, The method of claim 1 , wherein the model identifier is for uniquely identifying the target model.

10. The method of claim 1 , further comprising the step of the model training function node sending information of the target model to a model inference function node.

11. Before the step of the model training function node determining the FL server-side node, The method further includes the step of the model training function node determining that the target model should be obtained by federated learning; The step of the model training function node determining that the target model should be obtained by federated learning includes:

6. The method of claim 1, further comprising: if the model training function node determines that all or part of the training data for generating the target model cannot be obtained, determining that the target model should be obtained by federated learning.

12. a step of receiving a first request message sent by a model training function node by a federated learning FL server side node, the first request message being for triggering federated learning of the FL server side node to obtain a target model; The FL server node performs federated learning with the FL client node based on the first request message to obtain the target model; The FL server side node sends information of the target model to the model training function node; The FL server side node determines FL client nodes that will participate in federated learning; Including, The step of the FL server side node determining the FL client nodes to participate in federated learning includes: The FL server side node sends a node discovery request message to a network repository function network element to request FL client nodes to participate in federated learning; The FL server side node receives a response message sent by the network repository function network element, the response message including information of FL client nodes participating in federated learning; The node discovery request message A model information transmission method including at least one of interest time information and FL server side node type indication information.

13. The first request message including at least one of associative learning instruction information and a model identifier; the federated learning instruction information is for requesting the FL server side node to trigger federated learning to obtain a target model; The method of claim 12 , wherein the model identifier is for uniquely identifying the target model.

14. The method according to claim 12 or 13, wherein the first request message includes information of FL client nodes that will participate in federated learning.

15. The federation model information is: At least one of a model identifier, associative learning instruction information, a model file, and address information of the model file corresponding to the target model is included; the associative learning instruction information is for instructing that the target model is a model acquired by associative learning, the model identifier is for uniquely identifying the target model; The method comprises: The method of claim 12 or 13, further comprising the step of the FL server-side node obtaining the model identifier for the target model.

16. The method of claim 13 , wherein the model identifier is obtained by the model training function node for the target model.

17. a first determination module for determining a federated learning FL server side node; a first sending module for sending a first request message to the FL server side node to trigger federated learning of the FL server side node to obtain a target model; a receiving module for receiving information of a target model transmitted by the FL server side node; The first determination module is used for sending a node discovery request message to a network repository function network element to request network nodes to participate in federated learning training, and receiving a response message sent by the network repository function network element, wherein the response message includes information of the FL server side node; The node discovery request message A model information acquisition device including at least one of interest time information and FL server side node type instruction information.

18. a receiving module for receiving a first request message sent by a model training function node, the first request message being for triggering federated learning of a federated learning FL server side node to obtain a target model; a learning module for performing federated learning with an FL client node based on the first request message to obtain the target model; a transmission module for transmitting information of the target model to the model training function node; a determination module for determining FL client nodes to participate in federated learning; The determination module is used for sending a node discovery request message to a network repository function network element to request FL client nodes to participate in federated learning, and receiving a response message sent by the network repository function network element, wherein the response message includes information of FL client nodes that will participate in federated learning; The node discovery request message A model information transmitting device including at least one of interest time information and FL server side node type indication information.

Images