Associative learning methods, apparatus, communication devices, and readable storage media

Abstract

This application discloses a federated learning method, apparatus, communication device, and readable storage medium, belonging to the field of communication technologies. The federated learning method according to the embodiments of this application includes that a first communication device receives first information from a second communication device, where the first information includes at least one of second information for instructing whether the second communication device agrees to participate in federated learning, state information of the second communication device's current round of federated learning, and model performance information of the current round of federated learning, and determining whether the second communication device participates in the next round of federated learning based on the first information.

Description

Technical Field

[0001] (Cross-reference to Related Applications) This application claims the priority of Chinese Patent Application No. 202210815546.7 filed in China on July 8, 2022, and all the contents of the same application are incorporated herein by reference.

[0002] This application belongs to the field of communication technologies, and specifically relates to a federated learning method, apparatus, communication device, and readable storage medium.

Background Art

[0003] In related communication networks, in order to improve the model effect, the training of the model can be carried out based on federated learning. However, members participating in federated learning may, during the process of federated learning, for various reasons, such as not wanting to participate in federated learning because other more important tasks have arrived, or withdrawing from federated learning first because there are too many tasks to process, not want to participate in federated learning or become inappropriate for federated learning members. In such cases, how to reasonably select members participating in federated learning is an urgent problem that needs to be solved currently.

Summary of the Invention

Problems to be Solved by the Invention

[0004] Embodiments of this application provide a federated learning method, apparatus, communication device, and readable storage medium that can solve the problem of how to reasonably select members participating in federated learning.

Means for Solving the Problems

[0005] The first aspect provides a federated learning method, and this method is The first communication device receives first information from the second communication device, wherein the first information includes at least one of second information indicating whether the second communication device agrees to participate in federated learning, state information of the second communication device in this round of federated learning, and model performance information of the federated learning in this round. This includes the first communication device deciding, based on the first information, whether the second communication device will participate in the next round of federated learning.

[0006] The second aspect provides an associative learning method, which is: The second communication device determines the first information, wherein the first information includes at least one of the following: second information indicating whether the second communication device agrees to participate in federated learning; the state information of the second communication device's federated learning in this round; and the model performance information of the federated learning in this round. The second communication device transmits the first information to the first communication device, the first information being used by the first communication device to determine whether the second communication device will participate in the next round of federated learning.

[0007] A third aspect provides a federated learning device used in a first communication device, which device is A first receiving module for receiving first information from a second communication device, wherein the first information includes at least one of second information for indicating whether the second communication device agrees to participate in federated learning, state information of the second communication device's federated learning in this round, and model performance information of the federated learning in this round. The system includes a first decision module for determining whether the second communication device will participate in the next round of federated learning based on the first information.

[0008] A fourth aspect provides a federated learning device used in a second communication device, the device being A second decision module for determining first information, wherein the first information includes second information for indicating whether the second communication device agrees to participate in federated learning, state information of the second communication device in this round of federated learning, and model performance information of the federated learning in this round, A second transmitting module for transmitting the first information to a first communication device, the first information comprising the second transmitting module used by the first communication device to determine whether the second communication device will participate in the next round of federated learning.

[0009] A fifth aspect provides a communication device comprising a processor and a memory, the memory storing a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, it realizes a step of the method of the first aspect or a step of the method of the second aspect.

[0010] A sixth aspect provides a communication device comprising a processor and a communication interface, for example, if the communication device is a first communication device, the communication interface is used to receive first information from a second communication device, and the processor is used to determine, based on the first information, whether the second communication device will participate in the next round of federated learning; or if the communication device is a second communication device, the processor is used to determine first information, and the communication interface is used to transmit the first information to the first communication device, wherein the first information includes at least one of second information indicating whether the second communication device agrees to participate in the federated learning, state information of the second communication device in the current round of federated learning, and model performance information of the current round of federated learning.

[0011] A seventh aspect provides a communication system comprising a first communication device and a second communication device, wherein the first communication device may be used to perform the steps of the federated learning method described in the first aspect, and the second communication device may be used to perform the steps of the federated learning method described in the second aspect.

[0012] The eighth aspect provides a readable storage medium that stores a program or instruction, and when the program or instruction is executed by a processor, it realizes a step of the method according to the first aspect or a step of the method according to the second aspect.

[0013] The ninth aspect provides a chip comprising a processor and a communication interface, the communication interface being coupled with the processor, the processor being used to run a program or instructions, to implement a step of the method according to the first aspect, or to implement a step of the method according to the second aspect.

[0014] The tenth aspect provides a computer program / program product, the computer program / program product being stored in a storage medium, and the computer program / program product being executed by at least one processor to realize a step of the method of the first aspect or a step of the method of the second aspect. [Effects of the Invention]

[0015] In an embodiment of the present application, the first information is received from the second communication device, and based on the first information, it can be determined whether the second communication device participates in the next-round federated learning. The first information includes at least one of second information for indicating whether the second communication device agrees to participate in the federated learning, the status information of the second communication device in this round of federated learning, and the model performance information of this round of federated learning. Thereby, by combining the willingness, status information, and / or model performance of the second communication device, etc., and determining whether the second communication device participates in the next-round federated learning, a reasonable selection of members participating in the federated learning can be realized.

Brief Description of Drawings

[0016] [Figure 1] It is a block diagram of a wireless communication system to which an embodiment of the present application is applicable. [Figure 2] It is a schematic diagram of a neural network in an embodiment of the present application. [Figure 3] It is a schematic diagram of a neuron in an embodiment of the present application. [Figure 4] It is a flowchart of a federated learning method according to an embodiment of the present application. [Figure 5] It is a flowchart of another federated learning method according to an embodiment of the present application. [Figure 6] It is a schematic diagram of a federated learning process in an embodiment of the present application. [Figure 7] It is a schematic structural diagram of a federated learning device according to an embodiment of the present application. [Figure 8] It is a schematic structural diagram of another federated learning device according to an embodiment of the present application. [Figure 9] It is a schematic structural diagram of a communication device according to an embodiment of the present application. [Figure 10] It is a schematic structural diagram of a terminal according to an embodiment of the present application. [Figure 11] It is a schematic structural diagram of a network-side device according to an embodiment of the present application.

Modes for Carrying Out the Invention

[0017] The following clearly describes the technical concepts in the embodiments of this application, linking them to the drawings of the embodiments. Clearly, the embodiments described are only some, not all, embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are all within the scope of protection of this application.

[0018] The terms "first," "second," etc., used in the specification and claims of this application are intended to distinguish similar subjects and not to describe a specific order or sequence. It should be understood that these terms are interchangeable where appropriate, so that the embodiments of this application may be carried out in an order other than those illustrated or described herein, and that the subjects distinguished by "first" and "second" are generally of the same kind and do not limit the number of subjects; for example, the first subject may be one or more. Furthermore, "and / or" in the specification and claims indicates at least one of the connected subjects, and the letter " / " generally indicates that the preceding and succeeding related subjects are in an "or" relationship.

[0019] It should be noted that the technologies described in the embodiments of this application are not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, but are also applicable to 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 this application are always used interchangeably, and the technologies described may be used for the systems and radio technologies mentioned above, or for other systems and radio technologies. The following description describes a New Radio (NR) system for illustrative purposes, and uses NR terminology in most of the following descriptions, but these technologies are also applicable to applications other than NR system applications, such as sixth-generation (6) radio. th It may be applied to 6G (Generation 1) communication systems.

[0020] Figure 1 shows a block diagram of a wireless communication system to which an embodiment of this application can be applied. The wireless communication system includes a terminal 11 and a network-side device 12. Here, the terminal 11 is a mobile phone, tablet personal computer, laptop computer (or notebook computer), personal digital assistant (PDA), palmtop computer, netbook, ultra-mobile personal computer (UMPC), mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, vehicle user equipment (VUE), pedestrian user equipment (PUE), smart home (home appliances with wireless communication capabilities, such as refrigerators, televisions, washing machines or furniture), game console, personal computer (personal The terminal-side equipment may be a computer (PC), a deposit machine or self-service machine, and the wearable device includes smartwatches, smart wristbands, smart earphones, smart glasses, smart accessories (smart bracelets, smart hand chains, smart rings, smart necklaces, smart ankle bracelets, smart anklets, etc.), smart bands, smart clothing, etc. It should be noted that the terminal 11 in the embodiments of this application is not limited to a specific type. The network-side equipment 12 may include access network equipment or core network equipment, and the access network equipment may be called radio access network equipment, radio access network (RAN), radio access network function or radio access network unit.Access network equipment may include base stations, Wireless Local Area Networks (WLAN) access points, or WiFi nodes, and base stations may also be called node B, evolving node B (eNB), access point, base transceiver station (BTS), radio base station, radio transceiver, basic service set (BSS), extended service set (ESS), home B node, home evolving B node, transmitting and receiving point (TRP), or any other appropriate term in the art, and are not limited to specific technical terms as long as the same technical effect is achieved. For the purposes of this explanation, the embodiments of this application only use base stations in NR systems as examples and do not limit the specific types of base stations.The core network equipment includes Network Data Analytic Function (NWDAF), 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), and Binding Support Function. It may include, but is not limited to, at least one of the following: Function (BSF), Application Function (AF), etc. It should be noted that the embodiments of this application only describe core network equipment in an NR system as an example, and do not limit the specific type of core network equipment.

[0021] Selectively, in the embodiments of this application, the network data analysis function NWDAF may be divided into two network elements, for example, a Model Training Logical Network Element (MTLF) and an Analytics Logical Network Element (AnLF). Here, the Model Training Logical Network Element MTLF is mainly used to generate models and perform model training, and may be a central server in federative learning or a member (client) in federative learning. The Analytics Logical Network Element AnLF is mainly used to perform inference and generate predictive information or models, and may request a model from the MTLF, which may be generated by federative learning.

[0022] Selectively, the model in the embodiments of this application may be an artificial intelligence (AI) model. AI models can be implemented using various algorithmic methods, such as neural networks, decision trees, support vector machines, and Bayesian classifiers. This application uses a neural network as an example, but does not limit the specific type of AI module.

[0023] For example, a schematic diagram of a neural network may be shown as in Figure 2, where X1, X2, ... Xn are input values, Y is the output result, and each "〇" represents a single neuron, where the operation is performed and the result is passed on to the next layer. The input layer, hidden layer, and output layer, each consisting of many neurons, constitute a single neural network. The number of hidden layers and the number of neurons in each layer constitute the "network structure" of the neural network.

[0024] For example, a neural network consists of neurons, and a schematic diagram of a neuron may be shown as in Figure 3, where a1, a k …a K(i.e., X1, X2… shown in Figure 2) are inputs, w is the weight value (may also be called the multiplication coefficient), b is the bias (may also be called the addition coefficient), σ() is the activation function, z is the output value, and the corresponding calculation process is:

number

[0025] In actual use, an AI model is a file containing elements such as network structure and parameter information. A trained AI model can be directly reused by its framework platform, eliminating the need for repeated construction or training, and directly performing intelligent functions such as judgment and / or identification.

[0026] Federative learning aims to establish federative learning models based on distributed datasets. During the model training process, information related to the model can be exchanged between the parties (or in an encrypted form), but raw data cannot. This exchange does not expose any protected, private portions of the data on each training node.

[0027] Selectively, the federated learning according to the embodiments of this application is lateral federated learning. The essence of lateral federated learning is the linking of samples and is applicable to scenarios where the business types of the stakeholders are the same, but the customers they reach are different, i.e., there is a lot of overlap in features and little overlap in users. For example, the same service that provides services to users in different CN domains and RAN domains within a communication network (e.g., each UE, i.e., different samples), such as a mobility management (MM) service, a session management (SM) service, or some other service. By linking the same data features from different samples of stakeholders, lateral federation obtains a better model by increasing the number of training samples.

[0028] In the embodiments of this application, the server in federated learning (which may also be called a central server or organizer) may be a network element device in the network, such as an MTLF partitioned by an NWDAF. Members involved in federated learning (which may also be called clients) may be network element devices in the network, such as an MTLF partitioned by an NWDAF, or terminals. When performing federated learning, the server in federated learning may first select members to participate in the federated learning, for example by sending a request to a stored information network element such as an NRF to request the acquisition of capability information of each intelligent network element device such as an MTLF, matching whether they can participate in federated learning based on the capability information, and then sending information such as the federated learning initialization model to each selected member. After each member performs local model training, it feeds back intermediate results, such as gradients, to the server. The server then aggregates the received intermediate results and updates the global model. The steps of member selection - model distribution - local model training - intermediate result feedback - global model aggregation and update can be repeated multiple times, and model training can be stopped when the model converges or when other conditions are met.

[0029] In the following sections, the federated learning method, apparatus, communication equipment, and readable storage medium according to the embodiments of this application will be described in detail with reference to several embodiments and their application scenarios, while linking them to the drawings.

[0030] Referring to Figure 4, which is a flowchart of a federated learning method according to an embodiment of the present application, this method is used in a first communication device, which is specifically a server in federated learning and includes, but is not limited to, intelligent network element devices such as MTLF. As shown in Figure 4, this method includes the following steps.

[0031] Step 41: The first communication device receives the first information from the second communication device.

[0032] Step 42: Based on the information from the first communication device, the second communication device decides whether to participate in the next round of federated learning.

[0033] In this embodiment, the first information described above may include, but is not limited to, at least one of the following: second information for indicating whether the second communication device agrees to participate in federated learning; state information of the second communication device in this round of federated learning; and model performance information of the federated learning in this round. For example, this second information may optionally be motivation information for indicating whether the second communication device is willing to participate in federated learning.

[0034] Furthermore, the first piece of information described above may also include capability information for the second piece of communication equipment. For example, this capability information is the capability information after the model training of this round is completed, and includes, but is not limited to, whether or not it can become an associative learning participant (member), and accuracy information related to the training model. For example, after one local training session is completed, the capability information of a certain member may include that it can become an associative learning participant, that it has the ability to perform local training, and that the accuracy information related to the training model is X.

[0035] The second communication device mentioned above specifically refers to a member (client) device in federated learning, and may include, but is not limited to, terminals and intelligent network element devices such as MTLF.

[0036] In some embodiments, the first piece of information described above may be spontaneously reported by a second communication device (i.e., a member in federative learning), and by feeding it back to the server in federative learning, for example, along with the results of local training, the consumption of signaling and the number of interactions can be reduced.

[0037] In some embodiments, when a member in associative learning no longer wishes to participate—for example, if another, more important task arises—there are too many tasks to handle, and information indicating the member's unwillingness to participate in associative learning, i.e., information indicating their willingness to leave associative learning, can be fed back to the server in associative learning. This supports the selection of members in the associative learning process and enables rational selection of members involved in associative learning. On the other hand, if the member is willing to continue participating in associative learning, it is not necessary to provide feedback indicating their consent to participate, and in this case, the server tacitly acknowledges their willingness to continue participating. Furthermore, members in associative learning may directly instruct the server that they wish to participate in associative learning.

[0038] In some other embodiments, when the state of a member in associative learning changes (e.g., the load becomes heavier), the computational power required for local model training becomes insufficient, and the member is no longer suitable to be selected as a member to participate in the next round of associative learning. In such cases, the member in associative learning sends information about the state of that round of associative learning to the associative learning server, and the server decides whether or not to participate in the next round of associative learning. This supports member selection in the associative learning process, enables rational selection of members involved in associative learning, improves training efficiency, avoids the elimination of members whose state deteriorates (e.g., when a member whose state deteriorates does not provide feedback on results within a given time), and selects members who can deliver higher efficiency.

[0039] In some other embodiments, if a member's data in federated learning has already been trained multiple times or has already been incorporated into the global model of federated learning, this global model may become overfitted to the member's environment, making it unsuitable for selection as a member to participate in the next round of federated learning. Therefore, model convergence can be achieved more quickly by pausing training for this member for several rounds at this time. Thus, members in federated learning can send model performance information for the current round of federated learning to the server in federated learning, and the server can help select members in the federated learning process by deciding whether or not to participate in the next round of federated learning. This enables rational selection of members to participate in federated learning and improves training efficiency, for example, by avoiding the elimination of underperforming members (e.g., when underperforming members do not provide feedback on results within a given time) and by selecting members that can lead to higher efficiency.

[0040] Selectively, after receiving the first information described above, the first communication device may, based on this first information, select a third communication device to participate in the next round of federated learning. Unlike the second communication device, this third communication device is a new member (client) device specifically involved in the federated learning, and may include, but is not limited to, terminals and intelligent network element devices such as MTLF. For example, if, based on the received first information, it is determined that more member devices are no longer suitable to participate in the next round of federated learning, a new member can be selected to participate in the next round of federated learning, thereby ensuring that the federated learning proceeds smoothly.

[0041] In embodiments of this application, the state information may be used to describe the state information of a second communication device (i.e., a member in federated learning) after the completion of local training in this round of federated learning, and may include, but is not limited to, at least one of the following:

[0042] 1) Load information in the federated learning of the second communication device in this round.

[0043] In this embodiment, this load information may be understood as load status information, and may represent the load status of network elements, such as network functions (NF).

[0044] Selectively, this load information may include at least one of the following: average load information and peak load information. Average load information may be understood as the average load within the scope of associative learning in this round. For example, in a single local training session, a member's average load is 70%, and their peak load is 80%.

[0045] 2) Resource usage information for the second communication device in this round of federated learning.

[0046] In this embodiment, this resource usage information may be understood as resource usage status information.

[0047] Selectively, this resource usage information may include at least one of average resource usage information and peak resource usage information. Average resource usage information may be understood as the average resource usage within the scope of federated learning in this round.

[0048] For example, the resource usage endpoints corresponding to this resource usage information may include, but are not limited to, the Central Processing Unit (CPU), memory, magnetic disk, and graphics processing unit (GPU). This resource usage information may also include power consumption information.

[0049] For example, in a single local training session, a member's average resource usage is 60% CPU usage, 80% GPU usage, 70% memory usage (e.g., occupying 12GB, i.e., expressed numerically), and 40% disk usage, while this member's peak resource usage is 80% CPU usage, 100% GPU usage, 80% memory usage (e.g., occupying 14GB, i.e., expressed numerically), and 50% disk usage.

[0050] In the embodiments of this application, the model performance information is optionally model performance information before and / or after the start of local model training. The first model performance information after local model training is complete, This may include at least one of the following: second model performance information prior to the start of local model training.

[0051] Selectively, the above model performance information may include at least one of accuracy and Mean Absolute Error (MAE). Furthermore, it may also include, but is not limited to, at least one of the following: Precision, Recall rate, F1 score, Area Under Curve (AUC), Sum of Squares due to Error (SSE), sum variance, Mean Squared Error (MSE), variance, Root Mean Squared Error (RMSE), standard difference, and R-Squared coefficient.

[0052] In some embodiments, the first model performance information described above may include accuracy and mean absolute error MAE, etc. The second model performance information described above may include accuracy and mean absolute error MAE, etc.

[0053] To make it clear, the first model performance information described above is primarily used to describe the performance of the model on its local data after local model training is completed in this round of federated learning, and may include certain statistical parameters and the corresponding numerical values, such as the model accuracy and a specific value (e.g., 80%), and the mean absolute error MAE and its value (e.g., 0.1). The second model performance information described above is primarily used to describe the performance of the model on its local data before local model training begins in this round of federated learning, that is, after receiving the model, a statistical calculation of model performance must be performed once, and may include certain statistical calculation parameters and the corresponding numerical values, such as the model accuracy and a specific value (e.g., 70%), and the mean absolute error MAE and its value (e.g., 0.15).

[0054] It should be noted that accuracy is the percentage of correct predictions compared to the total number of predictions. During the model training phase, the dataset includes input data and labels (label data), which have a corresponding relationship. A set of input data corresponds to one or a set of labels, and the accuracy of the training is determined by comparing the predicted values ​​generated by the model with the labels corresponding to the current training. The mean absolute error (MAE) represents the average absolute error between the predicted values ​​and the true values, and the calculation method is as follows.

[0055]

number

[0056] In the embodiments of this application, whether or not the second communication device feeds back the first information may be determined by the first communication device. Selectively, the first communication device may transmit third information to the second communication device, which is used to identify that the second communication device needs to feed back the first information. On the other hand, if the first communication device does not transmit the third information, i.e., the second communication device does not receive the third information, the second communication device does not need to feed back the first information.

[0057] Selectively, the third piece of information is Information for identifying that the second communication device needs to feed back the second information (for example, this information is an identifier that needs to feed back the second information), Information to identify that the second communication device needs to feed back state information (for example, this information is an identifier that needs to feed back state information), Information for identifying when a second communication device needs to feed back model performance information (for example, this information may include, but is not limited to, an identifier for which model performance information needs to be fed back after local model training is completed, and / or an identifier for which model performance information needs to be fed back before local model training begins).

[0058] It should be noted that the information used to identify when the second communication device needs to feed back state information is primarily used to explain that the second communication device needs to feed back state information after the local model training for federative learning in this round is completed. It may also be specified that the specific state information could be at least one of the following: member load status (e.g., NF load), member resource usage status (e.g., resource usage including CPU, memory, disk, and / or GPU).

[0059] The information used to identify when the second communication device needs to feed back model performance information is primarily used to explain that the second communication device needs to feed back model performance information before the start and / or after the completion of local model training in this round of federated learning, and this model performance information includes the first and / or second model performance information described above.

[0060] Selectively transmitting the above third piece of information may include at least one of the following:

[0061] The first communication device transmits third information to the second communication device based on a pre-configured policy, where this pre-configured policy may specify when or under what circumstances the first communication device transmits third information to the second communication device, for example, after every five rounds of training, or after a second communication device has participated in five rounds of training. This pre-configured policy can not only indicate whether feedback from the second device is necessary, but also when or under what circumstances feedback should be requested. For example, if the pre-configured policy indicates that the second communication device needs to provide feedback on the first information, the first communication device may transmit third information to the second communication device; however, if the pre-configured policy indicates that the second communication device does not need to provide feedback on the first information, the first communication device will not transmit third information to the second communication device. This pre-configured policy may be predefined, stipulated by a protocol, etc.

[0062] The first communication device may transmit third information to the second communication device in response to the needs of the model training process based on associative learning. For example, if the first communication device expects the second communication device to decide whether to participate in the next round of associative learning based on the second communication device's willingness, state, and / or model performance, it may transmit third information to the second communication device; otherwise, it may not transmit third information to the second communication device. In other words, the first communication device may autonomously decide whether or not to transmit third information to the second communication device.

[0063] Selectively transmitting the third information described above may include transmitting a first request to a second communication device, the first request being used to request the second communication device to participate in federated learning, and the third information being carried in the first request. In this way, transmitting the third information with a first request to request the second communication device to participate in federated learning can reduce signaling consumption and the number of interactions.

[0064] In the embodiments of this application, when a first communication device receives multiple model performance information from multiple second communication devices, it may first combine the multiple model performance information to obtain a third model performance information, and then determine whether model training has been completed based on this third model performance information, for example, whether the model has converged. For example, if the third model performance information includes accuracy and this accuracy is higher than a preset threshold, it may be determined that model training has been completed, or if not, model training may be continued. Alternatively, if the third model performance information includes mean absolute error MAE and this MAE is lower than a preset threshold, it may be determined that model training has been completed, or if not, model training may be continued.

[0065] Selectively, the methods summarized above include, but are not limited to, calculating the average value for multiple model performance data, or calculating a weighted average value for multiple model performance data. When calculating a weighted average value, the weights may be determined by the first communication device, or, for example, pre-set weights or weights calculated by the user may be used.

[0066] In some embodiments, the first communication device aggregates multiple first model performance information (i.e., model performance information after local model training is completed) and determines whether model training has been completed based on the aggregated model performance information.

[0067] Furthermore, after obtaining the third model performance information, the first communication device may feed this third model performance information back to the model user to facilitate the model user's understanding of the model performance.

[0068] The above embodiment mainly describes the present application from the perspective of the first communication device (i.e., the server in federated learning), and the following describes the present application from the perspective of the second communication device (i.e., the member in federated learning).

[0069] Referring to Figure 5, Figure 5 is a flowchart of a federated learning method according to an embodiment of the present application, which is used in a second communication device, and this second communication device is specifically a member (client) in federated learning, and includes, but is not limited to, terminals and intelligent network element devices such as MTLF. As shown in Figure 5, this method includes the following steps.

[0070] Step 51: The second communication device determines the information from the first device.

[0071] Step 52: The second communication device transmits first information to the first communication device, which the first communication device uses to determine whether the second communication device will participate in the next round of federated learning.

[0072] In this embodiment, the first information described above may include, but is not limited to, at least one of the following: second information for instructing whether the second communication device agrees to participate in federated learning; state information of the second communication device in this round of federated learning; and model performance information of this round of federated learning.

[0073] The first communication device mentioned above is specifically a server in federated learning, and may include, but is not limited to, intelligent network element devices such as MTLF.

[0074] In some embodiments, the first piece of information described above may be spontaneously reported by a second communication device (i.e., a member in federative learning), and by feeding it back to the server in federative learning, for example, along with the results of local training, the consumption of signaling and the number of interactions can be reduced.

[0075] In the federated learning method according to the embodiment of this application, the first communication device transmits to the first communication device first information which includes second information to instruct the second communication device whether it agrees to participate in federated learning, the second communication device's state information for the current round of federated learning, first model performance information after the completion of local model training for the current round of federated learning, and second model performance information before the start of local model training for the current round of federated learning. By combining the willingness, state information, and / or model performance of the second communication device, the first communication device can decide whether the second communication device will participate in the next round of federated learning, thereby enabling a rational selection of members involved in federated learning, improving training efficiency, avoiding member dropouts (i.e., members who do not provide feedback on results within a predetermined time), and selecting members who can bring about higher efficiency.

[0076] In embodiments of this application, the state information may be used to describe the state information of a second communication device (i.e., a member in federated learning) after the completion of local training for this round of federated learning, and may include, but is not limited to, at least one of the following:

[0077] 1) Load information in the federated learning of the second communication device in this round.

[0078] In this embodiment, this load information may be understood as load status information, or it may represent the NF load status.

[0079] Selectively, this load information may include at least one of the following: average load information and peak load information. Average load information may be understood as the average load within the scope of associative learning in this round. For example, in a single local training session, a member's average load is 70%, and their peak load is 80%.

[0080] 2) Resource usage information for the second communication device in this round of federated learning.

[0081] Selectively, this resource usage information may include at least one of average resource usage information and peak resource usage information. Average resource usage information may be understood as the average resource usage within the scope of federated learning in this round.

[0082] For example, the resource users (e.g., resource usage) corresponding to this resource usage information may include, but are not limited to, the Central Processing Unit (CPU), memory, magnetic disk, and graphics processing unit (GPU). This resource usage information may also include power consumption information.

[0083] For example, in a single local training session, a member's average resource usage is 60% CPU usage, 80% GPU usage, 70% memory usage (e.g., occupying 12GB, i.e., expressed numerically), and 40% disk usage, while this member's peak resource usage is 80% CPU usage, 100% GPU usage, 80% memory usage (e.g., occupying 14GB, i.e., expressed numerically), and 50% disk usage.

[0084] In the embodiments of this application, the model performance information is optionally model performance information before and / or after the start of local model training. The first model performance information after local model training is complete, This may include at least one of the following: second model performance information prior to the start of local model training.

[0085] Selectively, the above model performance information may include at least one of accuracy and mean absolute error (MAE). For example, the first model performance information may include accuracy and mean absolute error (MAE). The second model performance information may include accuracy and mean absolute error (MAE).

[0086] In the embodiments of this application, whether or not the second communication device feeds back the first information may be determined by the first communication device. Determining the first information as described above may include first receiving third information from the first communication device, the third information being used to identify that the second communication device needs to feed back the first information, and then determining the first information based on the third information.

[0087] Selectively, the third piece of information is Information for identifying that the second communication device needs to feed back the second information (for example, this information is an identifier that needs to feed back the second information), Information to identify that the second communication device needs to feed back state information (for example, this information is an identifier that needs to feed back state information), Information for identifying when a second communication device needs to feed back model performance information (for example, this information may include, but is not limited to, an identifier for which model performance information needs to be fed back after local model training is completed, and / or an identifier for which model performance information needs to be fed back before local model training begins).

[0088] Selectively receiving third information from the first communication device may include receiving a first request from the first communication device, the first request being used to request the second communication device to participate in federated learning, and the third information is carried in the first request. In this way, by transmitting the third information with the first request to request the second communication device to participate in federated learning, the consumption of signaling and the number of interactions can be reduced.

[0089] The associated learning process in the embodiment of this application will be described below, with reference to Figure 6.

[0090] In the embodiments of this application, the federated learning server (server) is an NWDAF (e.g., MTLF), and the federated learning members (clients) are NWDAFs (e.g., MTLF), and as shown in Figure 6, the specific federated learning process includes the following:

[0091] Step 61: A federated learning consumer (e.g., NWDAF(AnLF)) sends a model request (e.g., Nnwdaf_MLModelProvision_Subscribe) to a federated learning server (e.g., NWDAF(MTLF)), which is used to request a model to complete its task. At this point, the server decides whether to trigger federated learning based on circumstances such as local configuration or the request from the federated learning consumer, and performs federated learning initialization and member selection.

[0092] Step 62: Once federated learning is triggered, the server may initialize and formulate policies for federated learning when selecting members, for example, specifying how many training rounds to collect state information once, and / or how many training rounds to collect model performance information.

[0093] Step 63: The server sends each member (client) a federated learning task request (e.g., Nnwdaf_MLModelTraining_Subscribe) to request participation in federated learning and to perform local training of the federated learning based on the global model and each member's local data. This task request may include a task identifier (e.g., analytic ID), model initialization information (e.g., training parameters), and information to identify whether state information needs to be fed back / model performance information (i.e., feedback requirement).

[0094] Here, the analytic ID is primarily used to indicate which task the model in question is used to perform. Model initialization information is used to describe the model and configuration information for the federated learning round. The descriptive model is the model itself, describing, for example, the structure of the model, such as the algorithm, architecture, parameters, and hyperparameters, or the model itself, such as the model file and the address information of the model file. Configuration information for the federated learning round refers to information such as the number of rounds of local training to be performed and the data types to be used in the local training process of the federated learning round. Information for identifying the need to feed back state information and model performance information may be found in the description of the above example and will not be explained further here.

[0095] Step 64: Members collect data and train local models by sending data acquisition requests (e.g., Ndccf_DataManagement_Subscribe / Nnf_EventExposure_Subscribe) to the region or data source to which they are located. Depending on the task, the network elements that provide the data will also differ, such as UPF, OAM, UDM, etc.

[0096] Step 65: The data source returns a response to the relevant member, which contains the requested data, and this response is, for example, Ndccf_DataManagement_Notify / Nnf_EventExposure_Notify.

[0097] Step 66: Each member uses the data obtained based on Steps 64 and 65 to train a local model, generate intermediate results, and feeds them back to the server in subsequent steps. The server then aggregates and updates the global model and uses the local data to analyze the model's performance.

[0098] For example, model performance analysis may involve calculating accuracy or MAE using the model after local training and local data. If the task request in step 63 includes identifier information that requires feedback of model performance information before local training, the member must perform statistical calculations of model performance before performing local training.

[0099] In one implementation, member NWDAF defines the model's local training accuracy as the number of times the model's prediction was correct divided by the total number of predictions; that is, the formula is Local Training Accuracy = Number of Correct Results ÷ Total Number of Predictions. Specifically, member NWDAF may set up a validation dataset to evaluate local training accuracy. This validation dataset includes model input data and true label data (label / ground truth). Member NWDAF inputs the input data into the trained model to obtain output data. Member NWDAF then compares the output data with the true label data to determine if they match, and uses the above formula to obtain the value of local training accuracy. Explanation: The concept of a correct prediction does not necessarily mean that the result perfectly matches the label data. If there was a certain difference between the two previously, but this difference is within an acceptable range, the prediction may be considered correct.

[0100] In one implementation, member NWDAF calculates the mean of the sum of squared point errors corresponding to the predicted data and label data (label values, raw data) to obtain MAE, and the following formula is used for local training:

number

number

[0101] Step 67: Each member, either voluntarily or in response to a request in Step 63, provides feedback to the server on the intermediate results of the completed local training and information such as motivation, state and / or model performance (the first information described above), for example, by a feedback message corresponding to a request message for the associative learning training process, which is optionally a notify message.

[0102] In one implementation, a member client may spontaneously provide feedback on model performance information if it discovers that its training status is good and, for example, its accuracy has reached a certain threshold (this threshold may be carried in the model initialization information in step 63, carried in the model request, or may be pre-acquired / configured). Alternatively, a member client may spontaneously provide feedback on its motivation, state, and / or model performance information in each round. The server can use the intermediate results to update the global model and use information such as motivation, state, and / or model performance to assist in member selection decisions during federated learning in the next round.

[0103] Step 68: The server aggregates the intermediate results and updates the global model based on the feedback received. Based on information such as the willingness to provide feedback, the status, and / or model performance, it determines whether the relevant members need to participate in the next round of federative learning. By aggregating the model performance information, the overall / global training status of the model can be obtained.

[0104] For example, after obtaining intermediate results from each client, the server can aggregate these intermediate results using the server's algorithm, such as averaging or weighted averaging, and then use these intermediate results to update the global model. Also, for example, the server can determine whether a client can participate in the next round of federative learning based on information such as motivation, state, and / or model performance provided by the client. For instance, if the client's motivation information indicates that it wants to exit federative learning, the server will select this client and not proceed to the next round of federative learning. Similarly, if the state information provided by the client indicates CPU usage of 90%, GPU usage of 100%, memory usage of 80% (e.g., 14GB, if expressed numerically), and magnetic disk usage of 50%, the server will determine whether this client is a good member to participate in federative learning. The server might consider that, when performing local training, the GPU is already fully utilized, potentially leading to longer training times or connection interruptions, and therefore may not select this client in the next round of member selection. Furthermore, for example, if the model performance feedback from this client is 98% accurate, but the model performance feedback from other clients is generally between 60% and 80%, the server might consider that the model is already overfitted in this client's environment and that training for this client needs to be paused, thus potentially not selecting this client in the next round of member selection.

[0105] It should be noted that aggregating model performance information to obtain the overall / global training status of the model means that the server collects model performance information from each client and generates a single global training status using methods such as averaging or weighted averaging. For example, if five clients participate in associative learning and feed back their model performance, and their accuracy is, for example, 70%, 72%, 75%, 68%, and 65%, the server can obtain the global training status by calculating the average of these accuracy values. That is, the accuracy of the global training status is (70% + 72% + 75% + 68% + 65%) / 5 = 70%. That is the case.

[0106] After the member re-selection is complete, steps 63 to 68 may be repeated until the model converges.

[0107] Step 69: After the associative learning model training is complete, the server provides feedback to the consumer (e.g., AnLF) regarding the trained model and overall / global model performance.

[0108] In the associative learning method according to the embodiment of this application, the execution unit may be an associative learning device. In the embodiment of this application, the associative learning device according to this application will be described as an example in which the associative learning device executes the associative learning method.

[0109] Referring to Figure 7, Figure 7 is a schematic diagram of the structure of a federated learning device according to an embodiment of the present application. This device is used in a first communication device, which specifically is a server in federated learning and includes, but is not limited to, intelligent network element devices such as MTLF. As shown in Figure 7, the federated learning device 70 is A first receiving module 71 for receiving first information from a second communication device, wherein the first information includes at least one of second information for indicating whether the second communication device agrees to participate in federated learning, status information of the second communication device's federated learning in this round, and model performance information of the federated learning in this round. The system includes a first decision module 72 for determining whether the second communication device will participate in the next round of federated learning based on the first information.

[0110] Selectively, the state information is Load information and, Includes at least one of the following: resource usage information.

[0111] Selectively, the load information includes at least one of average load information and peak load information. The resource usage information includes at least one of average resource usage information and peak resource usage information.

[0112] Selectively, the model performance information is The first model performance information after local model training is complete, This includes at least one of the second set of model performance information prior to the start of local model training.

[0113] Selectively, the model performance information includes at least one of accuracy, mean absolute error, precision, and mean squared error.

[0114] Selectively, the associative learning device 70, The system further includes a first transmitting module for transmitting third information to the second communication device, the third information being used to identify that the second communication device needs to feed back the first information.

[0115] Selectively, the third piece of information is Information for identifying that the second communication device needs to feed back the second information (for example, this information is an identifier that needs to feed back the second information), Information for identifying that the second communication device needs to provide feedback on its status, The second communication device includes at least one of the following: information for identifying that the second communication device needs to feed back model performance information.

[0116] Selectively, the first transmitting module specifically, Based on a pre-configured policy, the third piece of information is transmitted to the second communication device, It is used for at least one of the following: transmitting a third piece of information to the second communication device in response to the demands of a model training process based on associative learning.

[0117] Selectively, the first transmitting module is used to transmit a first request to the second communication device, the first request is used to request the second communication device to participate in joint learning, and the first request carries the third information.

[0118] Selectively, the associative learning device 70, The system further includes a processing module for determining whether model training has been completed when a first communication device receives multiple model performance information from multiple second communication devices, by combining the multiple model performance information to obtain a third model performance information, and based on the third model performance information.

[0119] Selectively, the associative learning device 70, The system further includes a feedback module for providing feedback on the third model performance information to the model user.

[0120] Selectively, the associative learning device 70, The system further includes a selection module for selecting a third communication device to participate in the next round of federated learning, based on the first information described above. Unlike the second communication device, this third communication device is specifically a new member (client) device involved in the federated learning, and may include, but is not limited to, terminals and intelligent network element devices such as MTLF. For example, if it is determined based on the received first information that more members are no longer suitable to participate in the next round of federated learning, new members can be selected to participate in the next round of federated learning to ensure that the federated learning proceeds smoothly.

[0121] The federated learning device 70 according to the embodiment of this application can realize each process realized by the embodiment of the method shown in Figure 4 and achieve the same technical effects, and to avoid repetition of the explanation, it will not be explained further here.

[0122] Referring to Figure 8, Figure 8 is a schematic diagram of the structure of a federated learning device according to an embodiment of this application, which is used in a second communication device, and this second communication device is specifically a member (client) in federated learning, and includes, but is not limited to, terminals and intelligent network element devices such as MTLF. As shown in Figure 8, the federated learning device 80 is, A second decision module 81 for determining first information, wherein the first information includes at least one of second information for indicating whether the second communication device agrees to participate in federated learning, state information of the second communication device in this round of federated learning, and model performance information of the federated learning in this round. A second transmitting module 82 for transmitting the first information to a first communication device, the first information being used by the first communication device to determine whether the second communication device will participate in the next round of federated learning.

[0123] Selectively, the state information is Load information and, Includes at least one of the following: resource usage information.

[0124] Selectively, the load information includes at least one of average load information and peak load information. The resource usage information includes at least one of average resource usage information and peak resource usage information.

[0125] Selectively, the model performance information is The first model performance information after local model training is complete, This includes at least one of the second set of model performance information prior to the start of local model training.

[0126] Selectively, the model performance information includes at least one of accuracy, mean absolute error, precision, and mean squared error.

[0127] Selectively, the associative learning device 80, The system includes a second receiving module for receiving third information from the first communication device, the third information being used to identify that the second communication device needs to feed back the first information. The second decision module 81 is specifically used to determine the first information based on the third information.

[0128] Selectively, the second receiving module is further used to receive a first request from the first communication device, the first request is used to request the second communication device to participate in joint learning, and the third information is carried in the first request.

[0129] The federated learning device 80 according to the embodiment of this application can implement each process realized by the embodiment of the method shown in Figure 5 and achieve the same technical effects, and to avoid repetition of the explanation, it will not be explained further here.

[0130] Selectively, as shown in Figure 9, embodiments of the present application further provide a communication device 90 comprising a processor 91 and a memory 92, the memory 92 storing a program or instruction that can be executed on the processor 91, for example, if this communication device 90 is a first communication device, when this program or instruction is executed by the processor 91, each step of the embodiment of the federated learning method shown in Figure 4 can be realized and the same technical effect can be achieved. If this communication device 90 is a second communication device, when this program or instruction is executed by the processor 91, each step of the embodiment of the federated learning method shown in Figure 5 can be realized and the same technical effect can be achieved, and to avoid repetition of the explanation, it will not be explained further here.

[0131] Embodiments of this application further provide a communication device comprising a processor and a communication interface, for example, if this communication device is a first communication device, the communication interface is used to receive first information from a second communication device, and the processor is used to determine, based on the first information, whether the second communication device will participate in the next round of federated learning; or if this communication device is a second communication device, the processor is used to determine first information, and the communication interface is used to transmit the first information to the first communication device, the first information comprising at least one of second information to indicate whether the second communication device agrees to participate in the federated learning, status information of the second communication device in the current round of federated learning, first model performance information after the local model training of the current round of federated learning is completed, and second model performance information before the local model training of the current round of federated learning is started. This embodiment corresponds to an embodiment of the above method, and each implementation process and implementation method of the embodiment of the above method can be applied to this embodiment and achieve the same technical effects.

[0132] Specifically, Figure 10 is a schematic diagram of the hardware structure that realizes the terminal of the embodiment of this application.

[0133] This terminal 1000 includes, but is not limited to, some of the following components: a radio frequency unit 1001, a network module 1002, an audio output unit 1003, an input unit 1004, a sensor 1005, a display unit 1006, a user input unit 1007, an interface unit 1008, a memory 1009, and a processor 1010.

[0134] As those skilled in the art will understand, the terminal 1000 may further include a power supply (e.g., a battery) to power each component, and the power supply may be logically connected to the processor 1010 by a power management system, thereby enabling functions such as charge / discharge management and power consumption management by the power management system. The terminal structure shown in Figure 10 does not constitute a limitation on the terminal, and the terminal may include more or fewer components than the number shown, or combinations of some components, or different arrangements of components, and will not be described further here.

[0135] It should be understood that, in the embodiments of this application, the input unit 1004 may include a graphics processing unit (GPU) 10041 and a microphone 10042, the graphics processor 10041 processing still images or video image data obtained by an image capture device (e.g., a camera) in video capture mode or image capture mode. The display unit 1006 may include a display panel 10061, which may be configured in the form of a liquid crystal display, organic light-emitting diodes, etc. The user input unit 1007 includes at least one of a touch panel 10071 and other input devices 10072. The touch panel 10071 is also called a touchscreen. The touch panel 10071 may include two parts: a touch detection device and a touch controller. The other input devices 10072 may include, but are not limited to, a physical keyboard, function keys (e.g., volume control buttons, switch buttons, etc.), a trackball, a mouse, or an operating lever, and will not be described further here.

[0136] In the embodiments of this application, the radio frequency unit 1001 can receive downlink data from network-side equipment and transmit it to the processor 1010 for processing, and the radio frequency unit 1001 can also transmit uplink data to network-side equipment. Generally, the radio frequency unit 1001 includes, but is not limited to, an antenna, amplifier, transceiver, coupler, low-noise amplifier, duplexer, etc.

[0137] Memory 1009 may be used to store software programs or instructions and various data. Memory 1009 may include a first storage area mainly for storing programs or instructions and a second storage area for storing data, wherein the first storage area can store an operating system, an application program or instructions necessary for at least one function (e.g., audio playback function, image playback function, etc.). Memory 1009 may include volatile memory or non-volatile memory, or memory 1009 may include both volatile and non-volatile memory. Here, non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (Erasable PROM, EPROM), electrically erasable programmable read-only memory (Electrically EPROM, EEPROM), or flash memory. Volatile memory may include Random Access Memory (RAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), Synch-link Dynamic Random Access Memory (SLDRAM), and Direct Rambus Random Access Memory (DRRAM). Memory 1009 in the embodiments of this application includes, but is not limited to, these and any other suitable types of memory.

[0138] The processor 1010 may include one or more processing units. Selectively, the processor 1010 integrates an application processor and a modem processor, where the application processor mainly handles operations related to the operating system, user interface, and application programs, and the modem processor mainly handles wireless communication signals, such as a baseband processor. To be clear, the above modem processor does not have to be integrated into the processor 1010.

[0139] Selectively, terminal 1000 may also be a member in federative learning, and processor 1010 is used to determine the first piece of information. The radio frequency unit 1001 is used to transmit first information to the server in federated learning, the first information being used by the server to determine whether terminal 1000 will participate in the next round of federated learning, and the first information includes at least one of the following: second information to indicate whether terminal 1000 agrees to participate in federated learning, status information of terminal 1000 in this round of federated learning, model performance information of this round of federated learning, etc.

[0140] The terminal 1000 according to the embodiment of this application can implement each process realized by the embodiment of the method shown in Figure 5 and achieve the same technical effects, and to avoid repetition of the explanation, it will not be explained further here.

[0141] Specifically, embodiments of this application further provide network-side equipment. As shown in Figure 11, this network-side equipment 110 includes a processor 111, a network interface 112, and memory 113. Here, the network interface 112 is, for example, a common public radio interface (CPRI).

[0142] Specifically, the network-side device 110 of the embodiment of this application further includes instructions or programs stored in memory 113 and operable on processor 111, the processor 111 calling instructions or programs in memory 113 and performing the same actions as those performed by the modules shown in Figures 7 and / or 8, and achieving the same technical effects, which are not described further here in order to avoid repetition of the description.

[0143] Embodiments of this application further provide a readable storage medium in which a program or instruction is stored, and when this program or instruction is executed by a processor, each process of the embodiment of the federated learning method described above can be realized and the same technical effects can be achieved, and to avoid repetition of the description, no further explanation is provided here.

[0144] Here, this processor is the processor in the terminal described in the above embodiment. This readable storage medium includes computer-readable storage media such as computer read-only memory ROM, random access memory RAM, magnetic disk or optical disk, etc.

[0145] Embodiments of this application further provide a chip comprising a processor and a communication interface, the communication interface being coupled with the processor, the processor running a program or instructions and used to implement each process of the embodiment of the federated learning method described above, and achieving the same technical effects, which are not described further here in order to avoid repetition of the description.

[0146] It should be understood that the chips referred to in the embodiments of this application may also be called system-level chips, system chips, chip systems, or system-on-a-chip, etc.

[0147] Embodiments of this application further provide a computer program / program product in which the computer program / program product is stored in a storage medium and executed by at least one processor to realize each process of the embodiment of the federated learning method described above and achieve the same technical effects, which will not be described further here in order to avoid repetition of the description.

[0148] Embodiments of this application further provide a communication system comprising a first communication device and a second communication device as described above, wherein the first communication device may be used to perform the steps of the federated learning method shown in Figure 4, and the second communication device may be used to perform the steps of the federated learning method shown in Figure 5.

[0149] It should be noted that, in this specification, the terms “include,” “incorporate,” or any other variation thereof are intended to cover the non-exclusive “include,” thereby including not only those elements but also other elements not explicitly listed, or elements specific to such process, method, article, or apparatus. Unless otherwise specified, an element limited by the phrase “includes one of…” is not excluded from the existence of other identical elements in a process, method, article, or apparatus containing that element. It should also be noted that the scope of methods and apparatus in embodiments of this application is not limited to performing functions in the order illustrated or discussed, but may include performing functions in a manner that is essentially simultaneous or in reverse order based on the functions involved, and methods described in a different procedure than those described, for example, may be performed, and various steps may be added, omitted, or combined. Furthermore, features described by reference to some examples may be combined with other examples.

[0150] As will be readily apparent to those skilled in the art from the above description of the embodiments, the methods of the above embodiments can be implemented in the form of software and a necessary general-purpose hardware platform. Of course, they may also be implemented in hardware, but in many cases the former is a more preferred embodiment. With this understanding in mind, the technical invention of this application may be embodied in substance or in part in relation to the art in the form of a computer software product, which is stored on a single storage medium (e.g., ROM / RAM, magnetic disk, optical disk) and contains some instructions for causing a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to perform the methods of each embodiment of this application.

[0151] The above describes embodiments of this application, accompanied by drawings; however, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can, by the suggestion of this application, make many forms, as long as they do not deviate from the spirit and claims of this application, and all of these fall within the scope of protection of this application.

Claims

1. Associative learning method, The first communication device receives first information from the second communication device, wherein the first information includes at least one of the following: state information of the second communication device's federated learning in this round, model performance information of the federated learning in this round, and motivation information for instructing the second communication device to exit the federated learning. The first communication device determines, based on the first information, whether the second communication device will participate in the next round of joint learning, The aforementioned method, The first communication device further includes transmitting third information to the second communication device, the third information being used to identify that the second communication device needs to feed back the first information. The third piece of information mentioned above is: Information for identifying that the second communication device needs to provide feedback on its status, A federated learning method comprising at least one of the following: information for identifying that the second communication device needs to feed back model performance information.

2. The aforementioned state information is, Load information and, The method according to claim 1, comprising at least one of the following: resource usage information.

3. The load information includes at least one of average load information and peak load information. The method according to claim 2, wherein the resource usage information includes at least one of average resource usage information and peak resource usage information.

4. The aforementioned model performance information is, The first model performance information after local model training is complete, The method according to claim 1, comprising at least one of the following: a second model performance information prior to the start of local model training.

5. The method according to claim 1, wherein the model performance information includes at least one of accuracy, mean absolute error, precision, and mean square error.

6. The first information further includes, The second communication device includes second information for indicating whether it agrees to participate in joint learning, The third piece of information mentioned above further states that The method according to claim 1, comprising information for identifying that the second communication device needs to feed back the second information.

7. Transmitting the third information to the second communication device is, The first communication device transmits third information to the second communication device based on a pre-configured policy, The method according to claim 1, further comprising at least one of the following: the first communication device transmitting a third piece of information to the second communication device in response to the demands of a model training process based on associative learning.

8. Transmitting the third information to the second communication device is, The method according to claim 1, comprising the first communication device transmitting a first request to the second communication device, the first request being used to request the second communication device to participate in joint learning, and the first request carrying the third information.

9. When the first communication device receives multiple model performance information from multiple second communication devices, the method is as follows: The first communication device combines multiple model performance information to obtain a third model performance information, The method according to claim 1, further comprising the first communication device determining whether model training has been completed based on the third model performance information.

10. After obtaining the third model performance information, the method proceeds as follows: The method according to claim 9, further comprising the first communication device feeding back the third model performance information to the model user.

11. After receiving the first information described above, the method shall The method according to claim 1, further comprising the first communication device selecting a third communication device to participate in the next round of federated learning based on the first information, wherein the third communication device is a new member device participating in the federated learning, unlike the second communication device.

12. The method according to claim 4, wherein the first model performance information includes accuracy or mean absolute error MAE, and the second model performance information includes accuracy or mean absolute error MAE.

13. Associative learning method, The second communication device determines the first information, wherein the first information includes at least one of the following: state information of the second communication device's federated learning in this round; model performance information of the federated learning in this round; and motivation information for instructing the second communication device to exit the federated learning. The second communication device transmits the first information to the first communication device, the first information being used by the first communication device to determine whether the second communication device will participate in the next round of federated learning, Determining the first piece of information mentioned above is The second communication device receives third information from the first communication device, and the third information is used to identify that the second communication device needs to feed back the first information. The second communication device determines the first information based on the third information, The third piece of information mentioned above is: Information for identifying that the second communication device needs to provide feedback on its status, A federated learning method comprising at least one of the following: information for identifying that the second communication device needs to feed back model performance information.

14. The aforementioned state information is, Load information and, The method according to claim 13, comprising at least one of the following: resource usage information.

15. The aforementioned model performance information is, The first model performance information after local model training is complete, The method according to claim 13, comprising at least one of the following: a second model performance information prior to the start of local model training.

16. It is an associative learning device, A first receiving module for receiving first information from a second communication device, wherein the first information includes at least one of the following: state information of the second communication device's federated learning in this round; model performance information of the federated learning in this round; and motivation information for instructing the second communication device to exit the federated learning. The first includes a first decision module for determining whether the second communication device will participate in the next round of federated learning based on the first information, The aforementioned associative learning device, The system further includes a first transmitting module for transmitting third information to the second communication device, the third information being used to identify that the second communication device needs to feed back the first information. The third piece of information mentioned above is: Information for identifying that the second communication device needs to provide feedback on its status, A federated learning device comprising at least one of the following: information for identifying that the second communication device needs to feed back model performance information.

17. It is an associative learning device, A second decision module for determining first information, wherein the first information includes at least one of the following: state information of the second communication device in this round of federated learning; model performance information of the federated learning in this round; and motivation information for instructing the second communication device to exit federated learning. A second transmitting module for transmitting the first information to a first communication device, wherein the first information includes the second transmitting module used by the first communication device to determine whether the second communication device will participate in the next round of federated learning. The aforementioned associative learning device, The system further includes a second receiving module for receiving third information from the first communication device, the third information being used to identify that the second communication device needs to feed back the first information. The second decision module is specifically used to determine the first information based on the third information. The third piece of information mentioned above is: Information for identifying that the second communication device needs to provide feedback on its status, A federated learning device comprising at least one of the following: information for identifying that the second communication device needs to feed back model performance information.

18. A communication device comprising a processor and a memory, wherein the memory stores a program or instruction that can be executed on the processor, and when the program or instruction is executed by the processor, it realizes a step of the federated learning method described in any one of claims 1 to 12, or a step of the federated learning method described in any one of claims 13 to 15.

19. A readable storage medium, the readable storage medium storing a program or instruction, and when the program or instruction is executed by a processor, the readable storage medium realizes the steps of the associative learning method described in any one of claims 1 to 12, or the steps of the associative learning method described in any one of claims 13 to 15.

Images