Node selection method and communication device

Abstract

Provided are a node selection method and a communication device, the method comprising: sending first information to N terminal devices, the first information comprising information of a first data set of a network device; second information from each terminal device in the N terminal devices is received, the second information comprises the statistical distance of each terminal device in the N terminal devices, and the statistical distance is used for indicating the distance between the second data set and the first data set of each terminal device in the N terminal devices; determining M terminal devices according to the second information; sending indication information, wherein the indication information is used for indicating M terminal devices; receiving a second data set from each terminal device in the M terminal devices; and performing model training according to the second data set of each terminal device in the M terminal devices. Based on the method, the network equipment can selectively sample the training data set, the convergence speed of model training is improved, and the communication overhead is reduced.

Description

Technical Field

[0001] This application relates to the field of communication technologies, and more particularly, to a node selection method and a communication device. Background Art

[0002] To improve the intelligence and automation levels of a network, inference models, such as artificial intelligence (AI) models and machine learning (ML) models, are applied to more and more technical fields. A model is usually obtained through training. In some application scenarios, the training of the model and the collection of the data set required for model training may be deployed on different devices. For example, multiple terminal devices may collect the training data set required for model training and send the collected training data set to a network device, and the network device performs model training to update the model. However, all terminal devices sending the training data set to the network device will introduce excessive communication losses and computational overheads.

[0003] In the existing solutions, the network device randomly samples the training data set collected by the terminal devices for training, but the convergence speed of the model is slow. Therefore, there is an urgent need for a new method that can enable the network device to selectively sample the training data set, improve the convergence speed of model training, and reduce communication overhead. Summary of the Invention

[0004] This application provides a node selection method and a communication device, which can enable the network device to selectively sample the training data set, improve the convergence speed of model training, and reduce communication overhead.

[0005] In a first aspect, a node selection method is provided. This method can be executed by a network device, or can also be executed by a component of the network device (such as a chip or a chip system or a circuit), and this is not limited. For ease of description, the following takes the execution by the network device as an example for illustration.

[0006] The method includes: sending first information to N terminal devices, where the first information includes information about a first data set of the network device; receiving second information from each of the N terminal devices, where the second information includes the statistical distance of each of the N terminal devices, and the statistical distance of each of the N terminal devices is used to indicate the distance between the second data set of each of the N terminal devices and the first data set; determining M terminal devices according to the second information; sending indication information, where the indication information is used to indicate the M terminal devices; receiving the second data set from each of the M terminal devices; and performing model training according to the second data set of each of the M terminal devices.

[0007] Wherein, N is an integer greater than 1; M is a positive integer, and M is less than or equal to N.

[0008] Based on the above technical solution, by considering the statistical distance between the training data sets collected by each of multiple terminal devices and the data set of the network device to select a terminal device, that is, to select the training data set collected by the terminal device, the network device can selectively sample the training data set, effectively utilize the data in the distributed environment, improve the convergence speed of model training, and reduce the communication overhead.

[0009] In combination with the first aspect, in some implementation manners of the first aspect, sending the indication information includes: sending a broadcast signal, where the broadcast signal includes first indication information for indicating a set of M terminal devices; or, sending second indication information to each of the M terminal devices, where the second indication information is used to indicate each terminal device.

[0010] Based on the above technical solution, the network device notifies the selected M terminal devices through a broadcast signal or by sending indication information separately, so that they know that they have been selected and can participate in model training, avoiding unnecessary communication. Since only the selected terminal devices need to participate in the transmission of data in model training, the system communication overhead is reduced.

[0011] In combination with the first aspect, in some implementation manners of the first aspect, determining M terminal devices according to the second information includes: determining M terminal devices according to the statistical distance ratio corresponding to each of the N terminal devices, where the statistical distance ratio is the ratio of the statistical distance of each of the N terminal devices to the maximum statistical distance among the statistical distances of the N terminal devices.

[0012] In some possible implementation manners, the M terminal devices are the M terminal devices with the largest statistical distance ratio among the N terminal devices.

[0013] Based on the above technical solution, the network device can select a terminal device according to the statistical distance between data sets, and select the terminal device with the largest statistical distance ratio, that is, the data set with a larger sampling error, which can accelerate model convergence.

[0014] In combination with the first aspect, in some implementation manners of the first aspect, the second information further includes the communication capabilities of each of the N terminal devices, and the communication capabilities include at least one of the following: the distance between each of the N terminal devices and the network device, or, the energy consumption per unit data volume of each of the N terminal devices.

[0015] In some possible implementation manners, selecting M terminal devices according to the second information includes: determining M terminal devices according to the statistical distance ratio and the communication capability ratio corresponding to each of the N terminal devices; where the statistical distance ratio is the ratio of the statistical distance of each of the N terminal devices to the maximum statistical distance among the statistical distances of the N terminal devices, and the communication capability ratio is the ratio of the communication capability of each of the N terminal devices to the maximum communication capability among the communication capabilities of the N terminal devices.

[0016] In some possible implementation manners, the M terminal devices are the M terminal devices with the largest target values among the N terminal devices, and the target value is the sum, or a weighted sum, of the statistical distance ratio and the communication capability ratio.

[0017] Based on the above technical solution, when the network device selects terminal devices, it comprehensively considers two factors of statistical distance and communication capability, and flexibly adjusts the weights of the two factors in the selection of terminal devices according to a weighted manner. This comprehensive selection method can balance the influence of statistical distance and communication capability, and can adapt to different requirements for communication capability and data similarity in different scenarios.

[0018] Combined with the first aspect, in some implementation manners of the first aspect, the information of the first data set is the first data set, and the statistical distance is the distance between the second data set of each of the N terminal devices and the first data set.

[0019] In some possible implementation manners, the information of the first data set is the statistical information of the first data set, and the statistical distance is the distance between the statistical information of the second data set of each of the N terminal devices and the statistical information of the first data set.

[0020] Optionally, the distance includes at least one of the following: Fréchet distance; maximum mean discrepancy; Earth Mover's (Wasserstein) distance; Kolmogorov-Smirnov distance.

[0021] Based on the above technical solution, obtaining the distance between data sets according to the information of the corresponding first data set, reasonably utilizing the computing capabilities of multiple terminal devices, reducing overhead, and having multiple options for the calculation method of the distance helps to consider the influence of different distance calculation methods on model training.

[0022] In a second aspect, a node selection method is provided. This method can be executed by a terminal device, or can also be executed by a component of the terminal device (such as a chip or a chip system or a circuit), and this is not limited. For the sake of description, the following takes the execution by the terminal device as an example for explanation.

[0023] The method includes: receiving first information from a network device, where the first information includes information on a first data set of the network device; sending second information, where the second information includes a statistical distance used to indicate the distance between a second data set of a terminal device and the first data set, and the second information is used to determine M terminal devices; receiving indication information used to indicate the M terminal devices; and in response to the indication information, sending the second data set for model training.

[0024] Where M is a positive integer.

[0025] Based on the above technical solution, the statistical distance helps to obtain the difference information between the data set of the terminal device and the data set of the network device, provides a measure of the statistical distance for the network device to select, can effectively utilize the data in the distributed environment, improve the convergence speed of model training, and reduce the communication overhead.

[0026] In combination with the second aspect, in some implementation manners of the second aspect, receiving the indication information includes: receiving a broadcast signal that includes first indication information used to indicate a set of M terminal devices; or receiving second indication information used to indicate the terminal device.

[0027] Based on the above technical solution, by receiving a broadcast signal or receiving separate indication information, the terminal device knows that it has been selected by the network device to participate in model training, and thus transmits the data set required for model training, avoiding unnecessary communication and thereby reducing the overall communication overhead.

[0028] In combination with the second aspect, in some implementation manners of the second aspect, the second information further includes communication capabilities, where the communication capabilities include at least one of the following: the distance between the terminal device and the network device, or the energy consumption per unit data volume of the terminal device.

[0029] Optionally, the terminal device obtains the distance from the network device, or obtains its own energy consumption per unit data volume.

[0030] Based on the above technical solution, by obtaining its own communication capabilities, the terminal device helps the network device to comprehensively consider factors such as communication efficiency and energy consumption when making a selection.

[0031] In combination with the second aspect, in some implementation manners of the second aspect, the information on the first data set is the first data set, and the statistical distance is the distance between the second data set and the first data set.

[0032] In some possible implementation manners, the information on the first data set is the statistical information of the first data set, and the statistical distance is the distance between the statistical information of the second data set and the statistical information of the first data set.

[0033] Optionally, the distance includes at least one of the following: Fréchet distance; maximum mean discrepancy; Earth Mover's (Wasserstein) distance; Kolmogorov-Smirnov distance.

[0034] Based on the above technical solutions, the terminal device can calculate the distance between data sets according to the information of the corresponding first data set, rationally utilize the computing capabilities of multiple terminal devices, reduce overhead, and there are multiple options for calculating the distance, which helps to consider the impact of different distance calculation methods on model training.

[0035] In a third aspect, the present application provides a communication device, which may include modules corresponding one by one to execute the methods / operations / steps / actions described in the first aspect or the second aspect. The module may be a hardware circuit, software, or a combination of hardware circuit and software. The communication device may be the first network element or the second network element, or a chip or circuit for the first network element or the second network element.

[0036] In one implementation, the communication device is a communication device. Exemplarily, the communication device may include a communication unit and / or a processing unit. The communication unit may be a transceiver, or an input / output interface; the processing unit may be at least one processor. Optionally, the transceiver may be a transceiver circuit. Optionally, the input / output interface may be an input / output circuit.

[0037] In another implementation, the device is a chip, chip system, or circuit for a communication device. When the device is a chip, chip system, or circuit for a terminal device, the communication unit may be an input / output interface, interface circuit, output circuit, input circuit, pin, or related circuit on the chip, chip system, or circuit; the processing unit may be at least one processor, processing circuit, or logic circuit, etc.

[0038] In one example, the communication device is a network device, or the communication device is a chip, chip system, or circuit provided in a network device, etc. In another example, the communication device is a terminal device, or the communication device is a chip, chip system, or circuit provided in a terminal device, etc.

[0039] In a fourth aspect, the present application provides a communication device, which includes a processor for executing a computer program or instruction stored in a memory to execute the method provided in the first aspect or the second aspect, or any one of its implementations. Optionally, the communication device further includes the memory. The communication device may be a network device or a terminal device, or a chip or circuit for a network device or a terminal device.

[0040] Fifth aspect, the present application provides a communication device, which includes a processor and a communication interface, and is configured to execute the method provided in the above first aspect or second aspect, or any one of their implementation manners. Exemplarily, the communication interface may be a transceiver, a hardware circuit, a bus, a module, a pin, etc., or other types of communication interfaces.

[0041] Sixth aspect, a chip or a chip system is provided, which includes a processor, and the processor is configured to run a program or an instruction so that the method in the first aspect or any possible implementation manner of the first aspect is implemented, and so that the method in the second aspect or any possible implementation manner of the second aspect is implemented. Optionally, the chip may further include a memory, and the memory is configured to store the program or the instruction. Optionally, the chip may further include the transceiver.

[0042] Seventh aspect, a computer-readable storage medium is provided, and the computer-readable storage medium includes instructions, and when the instructions are run by a processor, the method in the first aspect or any possible implementation manner of the first aspect is implemented, and the method in the second aspect or any possible implementation manner of the second aspect is implemented.

[0043] Eighth aspect, a computer program is provided, and when the computer program runs on a computer, the computer is caused to execute the above first aspect or second aspect, or the method provided in any implementation manner of the first aspect or second aspect.

[0044] Ninth aspect, a computer program product is provided, and the computer program product includes computer program code or instructions, and when the computer program code or instructions are run, the method in the first aspect or any possible implementation manner of the first aspect is implemented, and the method in the second aspect or any possible implementation manner of the second aspect is implemented.

[0045] Tenth aspect, a communication system is provided, which includes one or more of the foregoing network devices and terminal devices. Description of the Drawings

[0046] Figure 1 is a schematic diagram of a communication system provided by an embodiment of the present application.

[0047] Figure 2 is a schematic diagram of a fully connected neural network provided by an embodiment of the present application.

[0048] Figure 3 is a schematic diagram of a distributed learning provided by an embodiment of the present application.

[0049] Figure 4 is a schematic diagram of another distributed learning provided by an embodiment of the present application.

[0050] Figure 5 It is a node selection method provided by an embodiment of the present application.

[0051] Figure 6 It is another node selection method provided by an embodiment of the present application.

[0052] Figure 7 It is a schematic diagram of a communication device provided by an embodiment of the present application.

[0053] Figure 8 It is a schematic diagram of another communication device provided by an embodiment of the present application. Detailed implementation manners

[0054] Next, the technical solutions in the present application will be described with reference to the accompanying drawings.

[0055] The technical solutions of the embodiments of the present application can be applied to various communication systems, such as: Long Term Evolution (LTE) systems, LTE Frequency Division Duplex (FDD) systems, LTE Time Division Duplex (TDD), Worldwide Interoperability for Microwave Access (WiMAX) communication systems, 5th Generation (5G) systems or New Radio (NR) and future communication systems, Vehicle-to-Everything (V2X), where V2X can include Vehicle-to-Network (V2N), Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I), Vehicle-to-Pedestrian (V2P), etc., Long Term Evolution-Vehicle (LTE-V), vehicle networking, Machine-Type Communication (MTC), Internet of Things (IoT), Long Term Evolution-Machine (LTE-M), Machine-to-Machine (M2M), etc.

[0056] A device in a communication system can send a signal to another device or receive a signal from another device. The signal can include information, signaling, data, etc. Herein, the device can also be replaced with an entity, a network entity, a node, a communication device, a communication module, a network element, a communication node, etc. In this application, the description is given by taking the device as an example.

[0057] Figure 1 FIG. shows a schematic diagram of a communication system provided by an embodiment of this application, as Figure 1 shown, the communication system 100 may include one or more network devices. For example, Figure 1 the network devices 10, 20 shown, etc. The communication system 100 may also include one or more terminal devices. For example, Figure 1 the terminal devices 30, 40 shown, etc. Among them, the four terminal devices, namely the network device 10, the network device 20, the terminal device 30, and the terminal device 40, can communicate through a wireless link and thus exchange information.

[0058] In one implementation, the communication system may further include an AI entity. The network device may forward the data related to the AI model reported by the terminal device to the AI entity. The AI entity performs AI-related operations such as training dataset construction and model training, and provides the outputs of AI-related operations such as the trained AI model, model evaluation, and test results to the network device. In another implementation, the AI entity may also be located inside the network device 10 or the network device 20, that is, a module of the network device 10 or the network device 20.

[0059] In practical applications, a network device can serve one or more terminal devices at the same time. A terminal device can also be connected to one or more network devices at the same time. It can be understood that the network device and the terminal device can also be referred to as communication devices. The embodiments of this application do not limit the number of network devices and terminal devices included in the communication system.

[0060] Exemplarily, the network device may be a network-side device with wireless transceiver capabilities. The network device may be a device in a radio access network (RAN) that provides wireless communication capabilities for terminal devices, referred to as a RAN device. For example, the network device may be a base station, an evolved NodeB (eNodeB), a next-generation NodeB (gNB) in a 5G mobile communication system, a base station evolved by 3GPP subsequently, a transmission reception point (TRP), an access node in a WiFi system, a wireless relay node, a wireless backhaul node, etc. In communication systems using different radio access technologies (RATs), the names of devices with base station functions may vary. For example, in an LTE system, it may be referred to as an eNB or eNodeB, and in a 5G system or an NR system, it may be referred to as a gNB. The specific name of the base station in this application is not limited. The network device may include one or more co-located or non-co-located transmission reception points. Additionally, for example, the network device may include at least one of the following items: one or more central units (CUs), one or more distributed units (DUs), and one or more radio units (RUs). In different systems, the CU (or CU-CP and CU-UP), DU, or RU may also have different names, but those skilled in the art can understand their meanings. For example, the radio access network may also be an open radio access network (O-RAN) architecture. In an ORAN system, the CU may also be referred to as an O-CU (open CU), the DU may also be referred to as an O-DU, the CU-CP may also be referred to as an O-CU-CP, the CU-UP may also be referred to as an O-CU-UP, and the RU may also be referred to as an O-RU. Any of the CU (or CU-CP, CU-UP), DU, and RU units in this application may be implemented through software modules, hardware modules, or a combination of software modules and hardware modules. Exemplarily, the functions of the CU may be implemented by one entity or different entities. For example, the functions of the CU are further split, that is, the control plane and the user plane are separated and implemented by different entities, namely the control plane CU entity (i.e., the CU-CP entity) and the user plane CU entity (i.e., the CU-UP entity). The CU-CP entity and the CU-UP entity may be coupled to the DU to jointly complete the functions of the access network device.For example, the CU is responsible for processing non-real-time protocols and services, and implementing the functions of the radio resource control (RRC) layer and the packet data convergence protocol (PDCP) layer. The DU is responsible for processing physical layer protocols and real-time services, and implementing the functions of the radio link control (RLC) layer, the media access control (MAC) layer, and the physical (PHY) layer. In this way, some functions of the radio access network device can be implemented through multiple network function entities. These network function entities can be network elements in hardware devices, software functions running on dedicated hardware, or virtualized functions instantiated on a platform (such as a cloud platform). The first device may also include an active antenna unit (AAU). The AAU implements some physical layer processing functions, radio frequency processing, and related functions of the active antenna. Since the information of the RRC layer will ultimately become the information of the PHY layer, or is transformed from the information of the PHY layer, thus, in this architecture, high-layer signaling, such as RRC layer signaling, can also be considered to be sent by the DU, or sent by the DU + AAU. It can be understood that the network device can be a device including one or more of the CU node, the DU node, and the AAU node. In addition, the CU can be classified as a network device in the radio access network (RAN), or the CU can be classified as a network device in the core network (CN), and this application does not make a limitation on this. Another example is that in vehicle to everything (V2X) technology, the radio access network device can be a road side unit (RSU). Multiple radio access network devices in the communication system can be of the same type of base station, or different types of base stations. The base station can communicate with the terminal device, or communicate with the terminal device through a relay station. In the embodiments of this application, the device for implementing the functions of the network device can be the network device itself, or a device capable of supporting the network device to implement this function, such as a chip system or a combined device or component that can implement the functions of the radio access network device, and this device can be installed in the network device. In the embodiments of this application, the chip system can be composed of chips, or can include chips and other discrete devices.

[0061] Exemplarily, the terminal device may be a user-side device with wireless transceiver functions, which may be a fixed device, a mobile device, a handheld device (such as a mobile phone), a wearable device, a vehicle-mounted device, or a wireless device (such as a communication module, a modem, or a chip system, etc.) built into the above devices. The terminal device is used to connect people, objects, machines, etc., and can be widely used in various scenarios, such as: cellular communication, device-to-device (D2D) communication, V2X communication, machine-to-machine / machine-type communications (M2M / MTC) communication, Internet of Things, virtual reality (VR), augmented reality (AR), industrial control, self-driving, remote medical, smart grid, smart furniture, smart office, smart wear, smart transportation, smart city, drones, robots and other scenarios. Exemplarily, the terminal device may be a handheld terminal in cellular communication, a communication device in D2D, an IoT device in MTC, a surveillance camera in smart transportation and smart city, or a communication device on a drone, etc. The terminal device may sometimes be referred to as a user equipment (UE), a user terminal, a user device, a user unit, a user station, a terminal, an access terminal, an access station, a UE station, a remote station, a mobile device, or a wireless communication device, etc. The terminal device may also be a terminal device in an IoT system. IoT is an important part of the future development of information technology. Its main technical feature is to connect items to the network through communication technology, so as to realize an intelligent network of human-machine interconnection and thing-thing interconnection. In the embodiments of the present application, IoT technology can achieve massive connection, deep coverage, and power saving of the terminal through, for example, narrow band (NB) technology. In the embodiments of the present application, the device for implementing the functions of the terminal device may be the terminal device, or a device capable of supporting the terminal device to implement the functions, such as a chip system or a combined device or component that can implement the functions of the terminal device. This device may be installed in the terminal device.

[0062] The network device and the terminal device may be deployed on land, including indoor or outdoor, handheld or vehicle-mounted; they may also be deployed on the water surface; they may also be deployed on airplanes, balloons, and satellites in the air. In the embodiments of the present application, the scenarios where the network device and the terminal device are located are not limited.

[0063] Exemplarily, the communication system 100 may further include an Application Function (AF) network element, which is a control plane network function provided by the operator network and is used to provide application layer information; the communication system 100 may further include a Session Management Function (SMF) network element, which is a control plane network function provided by the operator network. In the embodiments of the present application, when the communication system 100 includes an AF network element and an SMF network element, the AF may send service-related information to the network device through the SMF.

[0064] It should be understood that Figure 1 The number and type of each device in the shown communication system are only for illustration, and the present application is not limited thereto. In practical applications, the communication system may further include more terminal devices, more network devices, more positioning devices, and may also include other network elements, such as core network devices, and / or network elements for implementing artificial intelligence functions.

[0065] It should also be understood that the division of the methods, situations, categories, and embodiments in the embodiments of the present application is only for the convenience of description and should not constitute a special limitation. The features in various methods, categories, situations, and embodiments may be combined without conflict.

[0066] It should also be understood that in various embodiments of the present application, the magnitude of the sequence numbers of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0067] It should be noted that in the embodiments of the present application, "pre-set", "pre-defined", "pre-configured", etc. can be implemented by pre-saving corresponding codes, tables, or other means that can be used to indicate relevant information in the device (for example, including terminal devices and network devices). The present application does not limit its specific implementation manner, such as the pre-configured information in the embodiments of the present application.

[0068] For clarity, some terms in the embodiments of the present application are explained below.

[0069] 1) Model: A function learned from data that can achieve specific functions / mappings. The model can be obtained based on technologies of artificial intelligence (AI) or machine learning (ML). Therefore, it can also be called an artificial intelligence / AI model, a machine learning / ML model, etc. Commonly used algorithms for generating AI / ML models include: supervised learning, unsupervised learning, and reinforcement learning. The corresponding models can be called supervised learning models, unsupervised learning models, and reinforcement learning models. For example, a supervised learning model can be a classification model, a prediction model, a regression model, etc., and an unsupervised learning model can be a clustering model. In addition, the model can also be obtained based on neural network (NN) technology, and such a model can also be called a neural network model, a deep learning model, etc.

[0070] 2) Training dataset: The data used for model training, validation, and testing in machine learning. The quantity and quality of the data will affect the effect of machine learning. The training data can either include the input of the AI model or include both the input and the target output of the AI model. Among them, the target output is the target value of the output of the AI model, and it can also be called the output true value, the output true value, the label, or the label sample.

[0071] It should be understood that the training dataset is a collection of training samples. Each training sample is an input to the neural network, and the training dataset is used for model training. The training dataset is one of the most important parts of machine learning. The training process of machine learning essentially involves learning certain features from the training dataset so that, under this training dataset, the difference between the output of the neural network and the ideal target value (i.e., the label or the output true value) is minimized. Usually, even with the same network structure, the weights and outputs of neural networks trained using different training datasets are different. Therefore, the composition and selection of the training dataset, to some extent, determine the performance of the trained neural network.

[0072] 3) Supervised learning: The goal of supervised learning is to, given a training dataset (containing multiple pairs of input and output), learn the mapping relationship between the input (data) and the output (label) in the training dataset. At the same time, it is hoped that this mapping relationship can also be applied to data outside the training dataset. The training dataset is a collection of correct input-output pairs.

[0073] Taking a fully connected neural network as an example, such as Figure 2As shown in the figure, it is a schematic diagram of a fully connected neural network provided by an embodiment of the present application. A fully connected neural network is also called a multilayer perceptron (MLP). An MLP includes an input layer (left side), an output layer (right side), and multiple hidden layers (in the middle). Each layer contains several nodes, called neurons. Among them, the neurons in adjacent layers are connected pairwise.

[0074] Considering the neurons in adjacent layers, the output h of the neurons in the next layer is the weighted sum of all the neurons x in the previous layer connected to it and passes through an activation function. It can be represented by a matrix as follows:

[0075] h = f(wx + b) (Formula 1)

[0076] Among them, w is the weight matrix, b is the bias vector, and f is the activation function.

[0077] Then the output of the neural network can be recursively expressed as:

[0078] y = f n (w n f n-1 (…)+b n ) (Formula 2)

[0079] Simply put, a neural network can be understood as a mapping relationship from an input data set to an output data set. Usually, neural networks are randomly initialized. The process of obtaining this mapping relationship from random w and b using existing data is called the training of the neural network.

[0080] The specific training method can use a loss function to evaluate the output result of the neural network, and backpropagate the error. By using the gradient descent method, w and b are iteratively optimized until the loss function reaches the minimum. The process of gradient descent can be expressed as:

[0081]

[0082] Among them, θ is the parameter to be optimized (such as w and b in the above text), L is the loss function, and η is the learning rate, which controls the step size of gradient descent.

[0083] The process of backpropagation can utilize the chain rule of partial derivatives, that is, the gradient of the previous layer's parameters can be recursively calculated from the gradient of the next layer's parameters. The formula can be expressed as:

[0084]

[0085] With the advent of the big data era, each device (including terminal devices and network devices) generates a huge amount of raw data in various forms every day, and this data will be born and exist in various corners of the world in the form of "isolated islands". Traditional centralized learning requires each edge device to uniformly transmit local data to the server at the central end, and then the server at the central end uses the collected data to train and learn the model. However, this architecture is gradually restricted by the following factors with the development of the times:

[0086] (1) Edge devices are widely distributed in various regions and corners of the world, and these devices continuously generate and accumulate a huge amount of raw data at a rapid speed. If the central end needs to collect the raw data from all edge devices, it will bring huge communication losses and computing power requirements.

[0087] (2) With the complication of actual scenarios in real life, more and more learning tasks require edge devices to be able to make timely and effective decisions and feedback. Traditional centralized learning involves a large amount of data upload, which will cause a relatively large delay, making it unable to meet the real-time requirements of actual task scenarios.

[0088] (3) Considering issues such as industry competition, user privacy and security, and complex administrative procedures, there will be increasing resistance and restrictions in centrally integrating data. Therefore, system deployment will increasingly tend to store data locally, and at the same time, the edge devices themselves will complete local computing.

[0089] It should be understood that the above central end can be the network device in the following text, and the edge device can be the terminal device in the following text. This application is not limited thereto.

[0090] The following introduces the technical solution provided by this application.

[0091] Figure 3 Shows a schematic diagram of a distributed learning provided by an embodiment of this application. As Figure 3 shown, in distributed learning, the network device can randomly select terminal devices to send the central data set. For example, send the central data set D (including data sets D a , D b , D c ) to these multiple terminal devices (such as the terminal device #a, terminal device #b, terminal device #c in Figure 3 ). These multiple terminal devices each use their local computing resources to train the model and will train the trained model W (including models W a , W b , W c)Uploaded to the network device. In this learning architecture, the network device has all the data sets, and the terminal device does not need to collect local data sets. This learning architecture utilizes the computing power of the terminal device to assist in model training and can offload the computing pressure of the network device.

[0092] It should be understood that the central data set is the data set collected by the network device for model training.

[0093] Although this distributed learning architecture offloads the computing pressure of the network device to a certain extent, there are also some disadvantages: First, after each training round ends, the terminal device needs to upload its locally trained model to the network device. This involves a large amount of model parameter transmission. Especially when the model is large or the network condition is poor, it may lead to significant communication overhead. Second, since the network device updates only based on the models uploaded by the terminal devices and does not use local data, this may result in a relatively slow model convergence speed. The network device may not be able to directly perceive the local data of each terminal device, thus affecting the global performance of the model.

[0094] Figure 4 Shows a schematic diagram of another distributed learning provided by an embodiment of the present application, as Figure 4 shown, the algorithm process is roughly as follows:

[0095] (1) The network device initializes the model to be trained and broadcasts it to all terminal devices, such as terminal device #1,..., terminal device #M,..., terminal device #N in the figure.

[0096] (2) Taking the Mth terminal device (terminal device #M) as an example, in the t-th round where t ∈ [1, T], terminal device #M trains the received global model M to obtain a local training result and reports it to the network device.

[0097] (3) The network device aggregates and collects the local training results from all (or part of) the terminal devices. Assuming that the set of terminals uploading local models in the t-th round is S t , the network device will perform weighted averaging with the number of samples of the corresponding terminal device as the weight to obtain a new global model. The specific update rule is where D′ M represents the number of data included in the data set D M . Then the network device broadcasts the latest version of the global model to all terminal devices for a new round of training.

[0098] ​(4) Repeat steps (2) and (3) until the model finally converges or the number of training rounds reaches the upper limit.

[0099] For the terminal device #M, in addition to reporting the local model to the network device it can also report the locally trained gradient. The network device averages the local gradients and updates the global model according to the direction of this average gradient.

[0100] Therefore, in the above distributed learning framework, the training dataset exists at the terminal device, that is, the terminal device collects the local dataset, performs local training, and reports the locally obtained results (model or gradient) to the network device. The network device itself has no dataset and is only responsible for fusing the training results of the terminal devices to obtain the global model and sending it to the terminal device. However, due to the periodic fusion of the entire model through the federated averaging algorithm, the convergence speed is slow and there are certain defects in the convergence performance; secondly, this distributed learning usually requires the network device to send the model to multiple terminal devices, which will cause too high communication overhead in the scenario of a large model scale and requires high computing, storage, and communication capabilities of the devices.

[0101] Therefore, how to select terminal devices to participate in training is a complex problem, which may affect the effect of model training. The embodiment of the present application provides a node selection method, which can improve the convergence speed and the utilization efficiency of data and computing power during model training. At the same time, in the solution provided by the present application, the interaction content between the network device and the terminal device is only the training dataset, which can significantly reduce the communication overhead when the number of model parameters is large and can also reduce the requirements for the computing, storage, and communication capabilities of the devices.

[0102] Figure 5 Fig. shows a node selection method 400 provided by an embodiment of the present application. As Figure 5 shown, the method 400 may include steps S410 to S470.

[0103] S410. The network device sends the first information to the terminal device; correspondingly, the terminal device receives the first information.

[0104] Exemplarily, the RRC layer of the network device sends the first information to the terminal device.

[0105] Wherein, the first information includes the information of the central dataset; the information of the central dataset may be the central dataset or the statistical information of the central dataset.

[0106] It should be understood that the central dataset is the dataset of the network device, and the dataset of the network device is used for model training. For the sake of convenience of description, the present application uses the central dataset to describe the dataset of the network device, and this is not limited thereto.

[0107] Optionally, the central dataset is a dataset obtained by the network device based on the updated model after the end of the previous round of model training.

[0108] Optionally, the statistical information of the central dataset includes at least one of the following: empirical distribution, statistical moments, compression, downsampling, etc.

[0109] Exemplarily, the empirical distribution refers to, for a given dataset, obtaining the probability distribution function of the dataset by normalizing the frequencies of the samples. It is the cumulative distribution function of the relative frequencies of each data point in the dataset and is used to describe the actually observed data distribution.

[0110] Exemplarily, statistical moments are statistics that describe the shape of the data distribution. Common moments include the mean, variance, skewness, and kurtosis, etc. The mean describes the central position of the dataset, the variance measures the degree of dispersion of the data, the skewness measures the symmetry of the distribution, and the kurtosis describes the peakedness or flatness of the distribution.

[0111] Exemplarily, compression can refer to the process of reducing the dimension of the dataset or extracting the main features. This can be achieved through various compression algorithms, principal component analysis (PCA), feature selection, etc. Compression helps to reduce the storage space of the data, lower the computational cost, and at the same time retain the key information about the importance of the data.

[0112] Exemplarily, downsampling refers to extracting a subset from the original dataset to reduce the number of data points. This helps to improve the computational efficiency when dealing with large-scale data, and in some cases, downsampling does not significantly lose the important information of the dataset. In statistical analysis and machine learning, downsampling can be used to quickly explore the data or reduce the data scale during the training process.

[0113] It should be understood that the network device can send the first information to multiple terminal devices in a broadcast form, and the number of the above terminal devices is not limited in this application.

[0114] Optionally, the first information further includes indication information for indicating the calculation method of the statistical distance.

[0115] It should be understood that the statistical distance is used to represent the distance between two datasets, and the calculation methods of the statistical distance include but are not limited to the Fréchet distance, maximum mean discrepancy, Earth Mover's (Wasserstein) distance, Kolmogorov - Smirnov distance.

[0116] To facilitate the understanding of the solution of this application, the calculation methods of the statistical distance are briefly introduced below.

[0117] (1) Fréchet distance

[0118] The Fréchet distance is a description of path space similarity. Specifically, given two curves or paths, each curve is regarded as a function defined on a common parameter space. The Fréchet distance takes into account the movement of the two paths in the parameter space, and when comparing them, it considers the consistency of both position and direction.

[0119] The calculation of the Fréchet distance involves the optimal matching between two functions. It can be regarded as placing two points between the two paths and then moving these two points along the paths so that they are always on the corresponding paths. The Fréchet distance is the maximum distance between these two points during the entire movement.

[0120] The specific calculation method can be expressed by the mathematical formula as:

[0121] F(P,Q)=inf αβ max t∈[0,1] d(P(α(t)),Q(β(t))) (Formula 5)

[0122] Where P and Q are two paths; α and β are the mappings parameterizing paths P and Q; t is a variable in the parameter space; d is the distance metric defined between points in the parameter space; inf represents taking the infimum over all possible α and β; max represents taking the maximum over the entire parameter space.

[0123] When using the Fréchet distance to measure the difference between datasets, we can regard each data point in the dataset as a point on the path, and the entire dataset is regarded as a path. That is to say, we can choose a distance metric defined between data points, which usually depends on the type and nature of the dataset. For example, the Euclidean distance, Manhattan distance, or other distance metrics suitable for the data type can be used; subsequently, parameterization can be introduced for the paths of each dataset by corresponding data points to points in the parameter space, which means that each data point will have a corresponding position in the parameter space, thus forming a path.

[0124] Exemplarily, assuming there are two datasets X and Y, datasets X and Y can be regarded as path X and path Y, then the formula for the Fréchet distance can be expressed as:

[0125]

[0126] Therefore, the Fréchet distance provides a quantitative way to measure the difference between two datasets. A smaller Fréchet distance indicates a higher similarity in the shapes of the datasets, while a larger Fréchet distance indicates a lower similarity in the shapes of the datasets.

[0127] (2) Maximum Mean Discrepancy

[0128] The Maximum Mean Discrepancy (MMD) is a method used to measure the distance between the distributions of two different but related random variables. MMD is widely applied in the field of machine learning, especially in the evaluation of transfer learning and generative models.

[0129] The core idea of MMD is to measure the difference between two distributions by comparing the sample means of the two distributions. Its calculation formula is usually based on the kernel method. A common definition is to calculate the MMD between two distributions P and Q by using the kernel function k:

[0130]

[0131] where f is the norm of the kernel function k in this space, and sup represents taking the supremum. The intuitive explanation of this formula is that MMD finds a maximized function f among all possible kernel functions k, such that the mean difference between P and Q under this function is the largest.

[0132] In transfer learning, MMD is often used as a loss function to measure the distribution difference between the source domain and the target domain. By minimizing MMD, the distributions of the two domains can be made closer, thereby improving the generalization performance of the model in the target domain. In addition, MMD can also be used to compare the distances between two datasets, especially when considering their distribution differences. The following is a simple example to illustrate how to use MMD to compare two datasets.

[0133] Suppose there are two datasets, denoted as dataset X and dataset Y, and each dataset contains multiple samples. For example, the sample points of dataset X are x, and the sample points of dataset Y are y. We can compare the distribution differences between these two datasets through the following steps:

[0134] First, select a suitable kernel function k. This usually depends on the nature of the data. For example, the Gaussian kernel is a good choice in many cases. Finally, use the selected kernel function k to calculate MMD, that is, calculate the MMD between the two datasets X and Y. The calculation of MMD can be achieved through the kernel method of sample means. The formula is as follows:

[0135]

[0136] Therefore, the smaller the value of MMD, the more similar the distributions between the two datasets. On the contrary, the larger the value of MMD, the greater the distribution difference between the two datasets.

[0137] (3) Earth Mover's (Wasserstein) Distance

[0138] The Wasserstein distance is used to compare the distances between two distributions and is particularly suitable for dealing with distributions with different shapes and qualities. Its core idea is the minimum cost required to transform one distribution into another, where the cost is weighted by the distance between the two distributions. One of the main advantages of the Wasserstein distance is that it is more robust in dealing with small changes and non-overlapping regions between distributions. The following is a simple example to illustrate how to use the Wasserstein distance to compare two datasets.

[0139] Suppose there are two datasets, denoted as dataset X and dataset Y, and each dataset contains multiple samples. For example, the sample points of dataset X are x, and the sample points of dataset Y are y. We can compare the distribution differences between these two datasets through the following steps:

[0140] First, regard each dataset as a probability distribution, where the probability of each sample occurring is equal to its frequency in the dataset; second, use the definition of the Wasserstein distance to calculate the distance between the two datasets. The calculation of the Wasserstein distance involves finding an optimal coupling in the sample space to minimize the total cost of transforming one distribution into another. The mathematical expression is as follows:

[0141]

[0142] where X and Y are the distributions of the two datasets, ∏(X, Y) is the set of all possible couplings between X and Y, and γ(x, y) is the cost function defined in the sample space.

[0143] Therefore, the value of the Wasserstein distance represents the minimum cost of transforming one distribution into another. A smaller Wasserstein distance usually indicates a higher similarity between the two datasets, while a larger distance indicates a greater difference between the two datasets.

[0144] (4) Kolmogorov-Smirnov (KS) distance

[0145] The KS distance is a statistic used to compare the differences between two probability distributions. It measures the maximum vertical distance between two cumulative distribution functions (CDFs). The KS distance is widely used to test whether two samples come from the same distribution, but it can also be used to compare the similarities of any two probability distributions. If the KS distance is zero, it means the two distributions are exactly the same; if the KS distance is large, it means the two distributions are quite different.

[0146] The calculation steps of the KS distance are as follows: First, for each sample set, sort it according to the numerical value, and then calculate the cumulative distribution function value corresponding to each value. This will generate a CDF for each sample set. Finally, for each sample value, calculate the absolute difference between the two CDFs and find the maximum difference. This maximum difference is the KS distance. Mathematically, the calculation of the KS distance can be expressed as:

[0147]

[0148] where F1(x) and F2(x) are the cumulative distribution functions of the two sample sets respectively, and x is the sample point.

[0149] When using the KS distance to measure the difference between data sets, we can regard the data sets as probability distributions, and indirectly use the KS distance to compare their similarities. Suppose there are two data sets, denoted as data set X and data set Y respectively, and each data set contains multiple samples. We can compare the distribution differences between these two data sets through the following steps: Obtain the sample data from data sets X and Y; regard each data set as a probability distribution and calculate their empirical distribution functions; use the KS distance Calculate the maximum difference between the two empirical distribution functions.

[0150] where F X (x) and F Y (x) are the empirical distribution functions of data sets X and Y respectively, and x is the sample point.

[0151] Therefore, when the KS distance is small, it indicates that the empirical distribution functions of the two data sets are relatively similar, that is, the difference between the two data sets is small; if the KS distance is large, it indicates that there is a large difference between the two data sets.

[0152] S420, the terminal device determines the second information according to the first information.

[0153] The second information includes the statistical distance, and the terminal device determines the statistical distance between the local data set of the terminal device and the central data set according to the first information.

[0154] Optionally, the terminal device calculates the distance between the local data set and the central data set according to the calculation method of the statistical distance indicated by the network device and the central data set.

[0155] Optionally, the terminal device calculates the distance between the statistical information of the local data set and the statistical information of the central data set according to the calculation method of the statistical distance indicated by the network device and the statistical information of the central data set.

[0156] Optionally, the statistical information includes at least one of the following: empirical distribution, statistical moment, compression, downsampling, etc.

[0157] It should be understood that the local data set is the data set collected by the terminal device.

[0158] Optionally, the second information further includes the communication capability of the terminal device.

[0159] It should be understood that the communication capability of the terminal device refers to various capabilities and performances that the terminal device possesses during communication, including the ability of the terminal device to connect to network devices, supported communication protocols, transmission rate, power consumption efficiency, location service, acquisition of sensor data, etc.

[0160] Exemplarily, the communication capability of the terminal device can be measured by the distance from the network device.

[0161] It should be understood that the Signal-to-Interference plus Noise Ratio (SINR) is an index used to measure the signal quality in wireless communication. It is the ratio of the received signal power to the received interference and noise power, and the mathematical expression is: Since the noise N and the interference I are relatively stable, so it can be known that to ensure that the receiving end (for example, the network device) receives a similar SINR, the transmission power P of the sending end (for example, the terminal device) tx and the square of the distance x 2 should remain constant. Therefore, we can use x 2 to represent the communication capability of the terminal device, that is, the distance between the terminal device and the network device can be used to represent the communication capability of the terminal device.

[0162] Exemplarily, the communication capability of the terminal device can also be measured by the communication quality of the terminal device or the energy consumption per unit data volume of the terminal device.

[0163] It should be understood that the communication capability can be measured by measuring indicators such as the actual communication quality, speed, and power consumption, that is, calculating the energy consumption per unit data volume according to the historical communication situation, because the historical communication situation of the terminal device provides the performance of the terminal device under different conditions.

[0164] Exemplarily, a mobile application can use the API of the mobile device (such as the Network StatsManager of Android) to monitor the data usage and data transmission speed of the application. Such measurements can be used to calculate the energy consumption required to send a unit data volume under specific network conditions.

[0165] S430, the terminal device sends the second information to the network device. Correspondingly, the network device receives the second information.

[0166] The second information includes a statistical distance.

[0167] Optionally, the second information further includes communication capabilities.

[0168] S440. The network device determines the terminal device according to the received second information.

[0169] Optionally, the second information includes a statistical distance, and the network device determines the terminal device according to the statistical distance.

[0170] Exemplarily, the network device receives the second information from each of the N terminal devices, and the second information includes the statistical distance of each of the N terminal devices. The network device determines M terminal devices with the largest statistical distance ratio according to the statistical distance ratio corresponding to each of the N terminal devices.

[0171] Wherein, the statistical distance ratio is the ratio of the statistical distance of each of the N terminal devices to the maximum statistical distance among the statistical distances of the N terminal devices.

[0172] Based on this solution, the network device can select the terminal device according to the statistical distance between the data sets, and select the terminal device with the largest statistical distance ratio, that is, the data set with a larger sampling error, to accelerate the model convergence.

[0173] Optionally, the second information further includes communication capabilities, and the network device determines the terminal device according to the statistical distance and communication capabilities.

[0174] Exemplarily, the network device receives the second information from each of the N terminal devices, and the second information includes the statistical distance of each of the N terminal devices and the communication capabilities of each of the N terminal devices. The network device determines M terminal devices with the largest target value according to the statistical distance ratio and communication capability ratio corresponding to each of the N terminal devices, and the target value is the sum, or weighted sum, of the statistical distance ratio and communication capability ratio.

[0175] Wherein, the statistical distance ratio is the ratio of the statistical distance of each of the N terminal devices to the maximum statistical distance among the statistical distances of the N terminal devices, and the communication capability ratio is the ratio of the communication capability of each of the N terminal devices to the maximum communication capability among the communication capabilities of the N terminal devices.

[0176] Based on this solution, the terminal device is selected by comprehensively considering two factors of communication capabilities and statistical distance, and the weights of the two factors in selecting the terminal device are flexibly adjusted according to the weighted method. This comprehensive selection method can balance the influence of statistical distance and communication capabilities, and can adapt to different requirements for communication capabilities and data similarity in different scenarios.

[0177] It should be noted that specific calculation examples in step S440 above will be introduced in detail in method 500 and will not be elaborated here.

[0178] S450, the network device sends indication information to the terminal device; correspondingly, the terminal device receives the indication information from the network device.

[0179] Optionally, the network device can broadcast a set of selected terminal devices (such as Figure 6 terminal device #1, terminal device #2,... terminal device #M in

[0180] Optionally, the identities of the selected terminal devices are included in the broadcast.

[0181] Optionally, the network device can send a selected signaling to each of the M selected terminal devices, and the selected signaling is used to indicate that the terminal device has been selected by the network device.

[0182] S460, the terminal device sends a data set to the network device; correspondingly, the network device receives the data set from the terminal device.

[0183] Exemplarily, the terminal device first needs to collect a local data set. That is to say, the local data set is a set of data generated, collected or stored locally by the terminal device, including user-generated information, sensor records, data generated by applications, etc. These data may cover various types, such as text, images, audio, sensor readings, etc., and the specific content depends on the application scenario and function of the terminal device. This application does not limit this. Secondly, the terminal device establishes a communication connection with the network device, encapsulates and sends the local data set to the network device to achieve data transmission and sharing.

[0184] S470, the network device performs model training according to the received data set.

[0185] Exemplarily, the network device receives the data set from each of the M selected terminal devices, and the network device performs model training according to the data set of each of the M terminal devices.

[0186] It should be understood that the network device trains the model according to the received data set, and this training can select a fixed number of rounds or stop when the model converges.

[0187] Optionally, the network device trains a new model or the model saved after the previous round of training according to the received data set to obtain an updated model.

[0188] Optionally, the central data set can be updated according to the updated model, and new data can be added, outdated data can be deleted, or existing data can be updated.

[0189] It should be understood that the data set described above (for example, the central data set or the data set of the terminal device) can be part or all of the data set itself, or it can be the indication information of the data set, and the present application does not limit this.

[0190] It should be noted that the updated central data set is used for the selection of the next round of terminal devices.

[0191] The solution of the present application should be applicable to the scenario of multi-round sampling. Assuming that K rounds of sampling are set, each round of sampling in the K rounds of sampling includes at least two terminal devices and one network device. For example, the c-th round of sampling in the K rounds of sampling includes N terminal devices and one network device. The i-th terminal device among the N terminal devices collects the local data set D i and calculates the statistical distance between the local data set D i and the central data set D c . The network device selects M terminal devices according to the statistical distance and communication capabilities corresponding to each of the N terminal devices, and sends indication information according to the selection. The indication information is used to indicate the selected M terminal devices. In response to the indication information, the M terminal devices send the collected local data sets to the network device. The network device trains the model for T rounds according to the data sets of each of the M terminal devices to obtain the trained and updated model W c .

[0192] For another example, the (c + 1)-th round of sampling in the K rounds of sampling includes H terminal devices and one network device. The i-th terminal device among the H terminal devices collects the local data set D i and calculates the statistical distance between the local data set D i and the central data set D c+1 . The network device selects Z terminal devices according to the statistical distance and communication capabilities corresponding to each of the H terminal devices, and sends indication information according to the selection. The indication information is used to indicate the selected Z terminal devices. In response to the indication information, the Z terminal devices send the collected local data sets to the network device. The network device trains the model for T rounds according to the data sets of each of the Z terminal devices to obtain the trained and updated model W c+1 .

[0193] It should be understood that H in the (c + 1)-th round of sampling is less than N in the c-th round of sampling; the central data set D c+1 in the (c + 1)-th round of sampling is the model W in the c-th round of sampling according to the network devicec The dataset to be updated.

[0194] The following takes the c-th sampling in the K-round sampling as an example to introduce the solution of this application.

[0195] Figure 6 FIG. shows a schematic diagram of a node selection method 500 provided by an embodiment of this application. The method 500 can be regarded as a specific expansion of the method 400, and may include but is not limited to steps S510 to S580.

[0196] S510, the network device sends first information to N terminal devices (such as Figure 6 terminal device #1, terminal device #2,..., terminal device #M,..., terminal device #N among them).

[0197] Correspondingly, each of the N terminal devices receives the first information from the network device.

[0198] Among them, the first information includes information of the central dataset; the information of the central dataset can be the central dataset or statistical information of the central dataset.

[0199] It should be understood that the central dataset is the dataset of the network device, and the dataset of the network device is used for model training. For the convenience of description, this application uses the central dataset to describe the dataset of the network device, and this is not limited.

[0200] Optionally, the central dataset is the dataset obtained by the network device according to the updated model after the end of the previous round of model training.

[0201] Exemplarily, the previous round of model is the model W obtained after the network device trains in the (c - 1)-th sampling c-1 , for the convenience of description, the central dataset of this round of sampling is denoted by D c , and the central dataset of the previous round of sampling is denoted by D c-1 .

[0202] It should be understood that during the (c - 1)-th sampling process, the network device sends first information to R terminal devices, and determines to select J terminal devices according to the statistical distance and communication capabilities sent by each of the R terminal devices. The J terminal devices send local datasets to the network device for training the model W in the (c - 2)-th sampling c-2 , after the model is updated, the model W c-1 is obtained, and according to the model W c-1 the central dataset D c-1 is updated to obtain the central dataset D for the c-th sampling c . Among them, R in the (c - 1)-th sampling is greater than N in the c-th sampling.

[0203] Optionally, the statistical information of the central data set includes at least one of the following: empirical distribution, statistical moments, compression, downsampling, etc.

[0204] For ease of description, hereinafter, the statistical information of the central data set is denoted as T(D c ), and the local data set collected by the terminal device #i is denoted as D i , where i = 1, 2, … M, … N. For example, the local data set collected by the terminal device #1 is D1, the local data set collected by the terminal device #2 is D2, …, the local data set collected by the terminal device #M is D M , …, and the local data set collected by the terminal device #N is D N , but the present application is not limited thereto.

[0205] Optionally, the first information further includes indication information for indicating the calculation method of the statistical distance, and the calculation methods of the statistical distance include: Fréchet distance, maximum mean discrepancy, Wasserstein distance, or Kolmogorov-Smirnov distance.

[0206] S520. Each of the N terminal devices respectively determines the statistical distances F1, F2, …, F M , …, F N according to the first information.

[0207] It should be understood that the terminal device #i calculates the statistical distance between the central data set D c and the local data set D i to obtain the statistical distance F i , where i = 1, 2, … M, … N.

[0208] Optionally, when the first information is the central data set D c , the terminal device #i calculates the statistical distance F c between the central data set D i and the local data set D i .

[0209] Exemplarily, when the first information indicates that the calculation method of the statistical distance is the Fréchet distance, the statistical distance F(1) is:

[0210]

[0211] Exemplarily, when the first information indicates that the calculation method of the statistical distance is the maximum mean discrepancy, the statistical distance F(2) is:

[0212]

[0213] Exemplarily, when the first information indicates that the calculation method of the statistical distance is the Wasserstein distance, the statistical distance F(3) is as follows:

[0214]

[0215] Exemplarily, when the first information indicates that the calculation method of the statistical distance is the Kolmogorov-Smirnov distance, the statistical distance F(4) is as follows:

[0216]

[0217] Wherein, and are respectively the empirical distributions of the central dataset D c and the local dataset D i , and x is the sample point.

[0218] Optionally, when the first information is the statistical information of the central dataset, the terminal device #i calculates the statistical distance between the statistical information T(D c ) of the central dataset and the statistical information T(D i ) of the local dataset.

[0219] It should be understood that the difference between the statistical information T(D c ) of the central dataset D c and the statistical information T(D i ) of the local dataset D i can also reflect the statistical distance between the two datasets.

[0220] Exemplarily, when the first information indicates that the calculation method of the statistical distance is the Fréchet distance, the statistical distance F(5) is as follows:

[0221]

[0222] Exemplarily, when the first information indicates that the calculation method of the statistical distance is the maximum mean discrepancy, the statistical distance F(6) is as follows:

[0223]

[0224] Exemplarily, when the first information indicates that the calculation method of the statistical distance is the Wasserstein distance, the statistical distance F(7) is as follows:

[0225]

[0226] Exemplarily, when the first information indicates that the calculation method of the statistical distance is the Kolmogorov-Smirnov distance, the statistical distance F(8) is as follows:

[0227]

[0228] Among them, and are the empirical distributions of the central dataset D c and the local dataset D i respectively, and x is the sample point.

[0229] Optionally, the terminal device can also calculate the statistical distance between the central dataset and the local dataset according to the configuration method of the statistical information.

[0230] Optionally, in S530, each of the N terminal devices obtains the communication capabilities C1, C2,..., C M ,..., C N .

[0231] Exemplarily, the terminal device #i obtains its own communication capability C i , where i = 1, 2,..., M,..., N.

[0232] Optionally, the square of the distance between the terminal device #i and the network device can be used to represent the communication capability C i .

[0233] It should be understood that SINR is an index used to measure the signal quality in wireless communication. It is the ratio of the received signal power to the received interference and noise power, and is mathematically expressed as: Since the noise N and the interference I are relatively stable, it can be known that to ensure that the receiving end receives a similar SINR, the transmission power P of the transmitting end tx and the square of the distance x 2 should remain constant. Therefore, we can use x 2 to represent the communication capability of the terminal device, that is, the distance between the terminal device and the network device can be used to represent the communication capability of the terminal device.

[0234] Optionally, the communication capability of the terminal device can also be measured by the communication quality of the terminal device or the energy consumption per unit data volume of the terminal device.

[0235] It should be understood that the communication capability can be measured by measuring indicators such as the quality, speed, and power consumption of actual communication, that is, calculating the energy consumption per unit data volume according to the historical communication situation, because the historical communication situation of the terminal device provides the performance of the node under different conditions.

[0236] In S540, each of the N terminal devices sends a second message to the network device, and the second message includes the statistical distance F i . Optionally, the second message further includes the communication capability C i .

[0237] Correspondingly, the network device receives second information from each of the N terminal devices.

[0238] S550, the network device determines M terminal devices according to the second information, where M ≤ N.

[0239] Optionally, the network device determines M terminal devices according to the statistical distance of each of the N terminal devices.

[0240] Exemplarily, the network determines the maximum statistical distance max among the statistical distances of the N terminal devices according to the statistical distance of each of the N terminal devices N F(D c ,D N ); then the network device traverses the N terminal devices to obtain the ratio of the statistical distance of each terminal device to the maximum statistical distance Select the statistical distance ratio The largest M terminal devices. Among them, D c is the central data set, D i is the data set of terminal device #i. For example, D N is the data set of terminal device #N.

[0241] Exemplarily, the network determines the maximum statistical distance max among the statistical distances of the N terminal devices according to the statistical distance of each of the N terminal devices N F(T(D c )); then the network device traverses the N terminal devices to obtain the ratio of the statistical distance of each terminal device to the maximum statistical distance Select the statistical distance ratio The largest M terminal devices. Among them, T(D c ) is the statistical information of the central data set D c of, T(D i ) is the statistical information of the data set D i of terminal device #i. For example, T(D N ) is the statistical information of the data set D N of terminal device #N.

[0242] Based on this solution, the network device can select terminal devices according to the statistical distance between data sets, and select the terminal devices with the largest statistical distance ratio, that is, the data sets with larger sampling errors, to accelerate model convergence.

[0243] Optionally, the network device selects M terminal devices according to the statistical distance and communication ability of each of the N terminal devices.

[0244] Exemplarily, the network first determines the maximum statistical distance max among the statistical distances of each of the N terminal devices based on the statistical distances of the N terminal devices N F(D c ,D N ), and the maximum communication capacity max among the communication capabilities of the N terminal devices N C N ; then the network device traverses the N terminal devices to obtain the ratio of the statistical distance of each terminal device to the maximum statistical distance and the ratio of the communication capacity of each terminal device among the N terminal devices to the maximum communication capacity Select the target value The M terminal devices with the largest values. Among them, D c is the central data set, D i is the data set of terminal device #i. For example, D N is the data set of terminal device #N; C i is the communication capacity of terminal device #i. For example, C N is the communication capacity of terminal device #N.

[0245] Exemplarily, the network first determines the maximum statistical distance max among the statistical distances of each of the N terminal devices based on the statistical distances of the N terminal devices N F(F(D c ),T(D N )) and the maximum communication capacity max among the communication capabilities of the N terminal devices N C N ; then the network device traverses the N terminal devices to obtain the ratio of the statistical distance of each terminal device to the maximum statistical distance and the ratio of the communication capacity of each terminal device among the N terminal devices to the maximum communication capacity Select the target value The M terminal devices with the largest values. Among them, T(D c ) is the statistical information of the central data set D c , T(D i ) is the statistical information of the data set D i of terminal device #i. For example, T(D N ) is the statistical information of the data set D N of terminal device #N; C i is the communication capacity of terminal device #i. For example, C N is the communication capacity of terminal device #N.

[0246] Based on this solution, when selecting terminal devices by comprehensively considering two factors of communication ability and statistical distance, the impacts of statistical distance and communication ability can be balanced, and different requirements for communication ability and data similarity in different scenarios can be met.

[0247] Optionally, when the network device selects terminal devices and has an obvious preference for communication ability or statistical distance, the weight between the two can be adjusted by the hyperparameter α.

[0248] Exemplarily, the network device selects the target value The M terminal devices with the largest value.

[0249] Exemplarily, the network device selects the target value The M terminal devices with the largest value.

[0250] Based on this solution, by adjusting the hyperparameter α, the trade-off between communication ability and statistical distance can be flexibly balanced to meet the optimization goals in specific requirements or scenarios. In practical applications, the optimal α value can be selected through methods such as cross-validation to achieve the best selection.

[0251] S560, the network device sends indication information to the terminal device. Correspondingly, the terminal device receives the indication information from the network device.

[0252] Optionally, the network device can broadcast a set of selected terminal devices (such as Figure 6 Terminal device #1, Terminal device #2,..., Terminal device #M).

[0253] Optionally, the broadcast contains the identifiers of the selected terminal devices.

[0254] Optionally, the network device can send a selected signaling to each of the M selected terminal devices, and the selected signaling is used to indicate that it has been selected by the network device.

[0255] Exemplarily, the network device sends a selected signaling to Terminal device #1, Terminal device #2,..., Terminal device #M respectively.

[0256] S570, each of the M selected terminal devices sends a data set to the network device.

[0257] Correspondingly, the network device receives the data sets from each of the M terminal devices.

[0258] Exemplarily, each of the M terminal devices sends a data set (D1, D2,..., D M ) to the network device.

[0259] S580, the network device trains a model according to the received data sets (D1, D2, …, D M ).

[0260] Optionally, the network device may determine a fixed number of training rounds T, perform model training, and obtain an updated model W c ;

[0261] Optionally, update the central data set according to the model W c , and the updated central data set can be used for the next round of sampling.

[0262] After the network device obtains the training data set required for the model, it can train the model.

[0263] Optionally, if the model has not converged and multiple operations need to be performed, the process shown Figure 6 can be repeated.

[0264] Exemplarily, the network device performs T times of model training in the c-th round of sampling in K rounds of sampling to obtain the model W c , repeat Figure 6 the process shown, and the network device can perform T times of model training again in the (c + 1)-th round of sampling to obtain the model W c+1 , where the value of K can be preset, or sampling stops until the model converges.

[0265] It should be understood that the central data set for the next round of sampling, that is, the (c + 1)-th round of sampling, is the central data set updated after the c-th round of sampling ends.

[0266] Optionally, for the (c + 1)-th round of sampling, the number of terminal devices can be reduced. For example, H terminal devices can be selected for the (c + 1)-th round of sampling, and H is less than N.

[0267] In the solution provided in the embodiments of the present application, after the network device performs K*T times of model training in K rounds of sampling, the model obtained after completing K*T times of model training can improve the convergence speed during the model training process, and the solution of the present application trains the model in multiple rounds, which can reduce the requirements for the computing, storage, and communication capabilities of the device.

[0268] It can be understood that when operations of the model, such as training and / or inference, are not performed on the network device, the above steps S470 and S580 can be replaced by a model operation network element, and the method further includes the network device sending information related to the operations of the model or reporting information to the model operation network element.

[0269] It should be understood that the data sets described above (for example, the central data set, or the data set of the terminal device) can be part or all of the data set itself, or can be indication information of the data set, and the present application does not limit this.

[0270] The method for node selection provided in this application has been described in detail above. As described above, the method provided in this application can be applied to the AI scenario, in which case the above model is an AI model. The method of this application can also be applied to non-AI scenarios. For example, if a network device needs to select a data set conditionally, the method of this application can be used.

[0271] As described above in conjunction with Figures 5 to 6 the embodiments of the node selection method provided in this application have been described. Next, in conjunction with Figure 7 the embodiments of the device provided in this application will be described. It should be understood that the device embodiments and the method embodiments correspond to each other, and similar descriptions can refer to the method embodiments.

[0272] Figure 7 FIG. shows a schematic diagram of a device 700 provided in an embodiment of this application. The device 700 may include a transceiver unit 710, a storage unit 720, and a processing unit 730. The transceiver unit 710 is configured to receive or send information and / or data. The transceiver unit 710 may also be referred to as a communication interface or a communication unit. The storage unit 720 is configured to implement a corresponding storage function and store corresponding information and / or data. The processing unit 730 is configured to perform data processing so that the device 700 implements the foregoing node selection method.

[0273] In a possible implementation, the device 700 may only include the transceiver unit 710 and the processing unit 730.

[0274] As a design, the device 700 may perform the actions performed by the network device in the above method embodiments.

[0275] The device 700 includes: a transceiver unit 710 and a processing unit 730. The transceiver unit 710 is configured to: send first information, where the first information includes information about a first data set of the network device; receive second information from each of the N terminal devices, where the second information includes the statistical distance of each of the N terminal devices, and the statistical distance of each of the N terminal devices is used to indicate the distance between the second data set of each of the N terminal devices and the first data set. The processing unit 730 is configured to determine M terminal devices according to the second information.

[0276] In a possible implementation, the transceiver unit 710 is further configured to send indication information, where the indication information is used to indicate the M terminal devices; receive second data sets from each of the M terminal devices. The processing unit 730 is further configured to perform model training according to the second data sets of each of the M terminal devices.

[0277] In a possible implementation, the transceiver unit 710 is specifically configured to send a broadcast signal, where the broadcast signal includes first indication information for indicating a set of M terminal devices; or, send second indication information to each of the M terminal devices, where the second indication information is used to indicate each terminal device.

[0278] In a possible implementation, the processing unit 730 is specifically configured to determine M terminal devices according to the statistical distance ratio corresponding to each of the N terminal devices, where the statistical distance ratio is the ratio of the statistical distance of each of the N terminal devices to the maximum statistical distance among the statistical distances of the N terminal devices.

[0279] Among them, the M terminal devices are the M terminal devices with the largest statistical distance ratio among the N terminal devices.

[0280] In a possible implementation, the processing unit 730 is specifically configured to determine M terminal devices according to the statistical distance ratio and communication ability ratio corresponding to each of the N terminal devices; where the statistical distance ratio is the ratio of the statistical distance of each of the N terminal devices to the maximum statistical distance among the statistical distances of the N terminal devices, and the communication ability ratio is the ratio of the communication ability of each of the N terminal devices to the maximum communication ability among the communication abilities of the N terminal devices.

[0281] Among them, the M terminal devices are the M terminal devices with the largest target value among the N terminal devices, and the target value is the sum, or weighted sum, of the statistical distance ratio and the communication ability ratio.

[0282] As a design, the apparatus 700 can perform the actions executed by the terminal device in the above method embodiments.

[0283] The apparatus 700 includes: a transceiver unit 710 and a processing unit 730. The transceiver unit 710 is configured to: receive first information from a network device, where the first information includes information about a first data set of the network device; send second information, where the second information includes a statistical distance for indicating the distance between a second data set of a terminal device and the first data set, and the second information is used to determine M terminal devices; receive indication information for indicating the M terminal devices; and in response to the indication information, send the second data set for model training.

[0284] In a possible implementation, the transceiver unit 710 is specifically configured to receive a broadcast signal, where the broadcast signal includes first indication information for indicating a set of M terminal devices; or, receive second indication information for indicating the terminal device.

[0285] In one possible implementation, the processing unit 730 is configured to determine the statistical distance.

[0286] Optionally, the information of the first data set is the first data set, and the statistical distance is the distance between the second data set and the first data set.

[0287] Optionally, the information of the first data set is the statistical information of the first data set, and the statistical distance is the distance between the statistical information of the second data set and the statistical information of the first data set.

[0288] Figure 8 FIG. shows a schematic diagram of another device 800 provided by an embodiment of the present application.

[0289] The device 800 may include a memory 810, a processor 820, and a communication interface 830. Among them, the memory 810, the processor 820, and the communication interface 830 are connected through an internal connection path. The memory 810 is used to store instructions, and the processor 820 is used to execute the instructions stored in the memory 810 to control the communication interface 830 to obtain information, or to enable the device 800 to implement the foregoing model training method. Optionally, the memory 810 can be coupled to the processor 820 through an interface or integrated with the processor 820.

[0290] In one possible implementation, the device 800 may only include the processor 820 and the communication interface 830.

[0291] It should be noted that the above communication interface 830 uses a transceiver device such as, but not limited to, a transceiver. The above communication interface 830 may further include an input / output interface.

[0292] The processor 820 stores one or more computer programs, and the one or more computer programs include instructions. When the instructions are run by the processor 820, the device 800 is caused to execute the model training method in the above embodiments.

[0293] In the implementation process, the steps of the above method can be completed by the integrated logic circuit in the hardware of the processor 820 or the instructions in the form of software. The method disclosed in combination with the embodiments of the present application can be directly embodied as being executed by the hardware processor, or executed by a combination of the hardware and software modules in the processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory 810, and the processor 820 reads the information in the memory 1110 and combines its hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0294] In a possible implementation, the apparatus 800 may only include a processor 820 and a communication interface 830, and does not include a memory 810.

[0295] Optionally, Figure 8 the communication interface 830 in Figure 7 may implement the transceiver unit 710 in Figure 8 and the processor 820 in Figure 7 may implement the processing unit 730 in

[0296] Embodiments of the present application further provide a computer-readable storage medium storing program codes, which when run on a computer, cause the computer to execute any of the above Figure 5 or Figure 6 methods.

[0297] Embodiments of the present application further provide a computer program product including a computer program, which when run, causes a computer to execute any of the above Figure 5 or Figure 6 methods.

[0298] Embodiments of the present application further provide a chip or a chip system, including: a circuit configured to execute any of the above Figure 5 or Figure 6 methods.

[0299] Embodiments of the present application further provide a communication system, including: a first device and a second device, where the first device is configured to execute the actions / steps performed by the network device in Figure 5 or Figure 6 ; and the second device is configured to execute the actions / steps performed by the terminal device in Figure 5 or Figure 6

[0300] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present application.

[0301] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0302] ​In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections between each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0303] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0304] In addition, in each embodiment of the present application, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0305] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the essence of the technical solution of the present application, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0306] As described above, the above are only the specific implementation manners of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A node selection method, characterized in that, Including: Sending first information to N terminal devices, where the first information includes information on a first data set of a network device, and N is an integer greater than 1; Receiving second information from each of the N terminal devices, where the second information includes the statistical distance of each of the N terminal devices, and the statistical distance of each of the N terminal devices is used to indicate the distance between the second data set of each of the N terminal devices and the first data set; Determining M terminal devices according to the second information, where M is a positive integer and M is less than or equal to N; Sending indication information, where the indication information is used to indicate the M terminal devices; Receiving the second data set from each of the M terminal devices; Performing model training according to the second data set of each of the M terminal devices.

2. The method according to claim 1, wherein The sending of the indication information includes: Sending a broadcast signal, where the broadcast signal includes first indication information, and the first indication information is used to indicate the set of the M terminal devices; or, Sending second indication information to each of the M terminal devices, where the second indication information is used to indicate each of the terminal devices.

3. The method according to claim 1 or 2, characterized in that, The determining of the M terminal devices according to the second information includes: Determining the M terminal devices according to the statistical distance ratio corresponding to each of the N terminal devices, where the statistical distance ratio is the ratio of the statistical distance of each of the N terminal devices to the maximum statistical distance among the statistical distances of the N terminal devices.

4. The method according to claim 3, characterized in that The M terminal devices are the M terminal devices with the largest statistical distance ratio among the N terminal devices.

5. The method according to any one of claims 1 to 4, characterized in that, The second information further includes the communication capability of each of the N terminal devices, and the communication capability includes at least one of the following: the distance between each of the N terminal devices and the network device, the energy consumption per unit data volume of each of the N terminal devices.

6. The method according to claim 5, characterized in that, The determining of the M terminal devices according to the second information includes: Determining M terminal devices according to the statistical distance ratio and the communication capability ratio corresponding to each of the N terminal devices; where the statistical distance ratio is the ratio of the statistical distance of each of the N terminal devices to the maximum statistical distance among the statistical distances of the N terminal devices, and the communication capability ratio is the ratio of the communication capability of each of the N terminal devices to the maximum communication capability among the communication capabilities of the N terminal devices.

7. The method according to claim 6, characterized in that The M terminal devices are the M terminal devices with the largest target value among the N terminal devices, and the target value is the sum, or weighted sum, of the statistical distance ratio and the communication capability ratio.

8. The method according to any one of claims 1 to 7, characterized in that The information of the first data set is the first data set, and the statistical distance is the distance between the second data set of each of the N terminal devices and the first data set.

9. The method according to any one of claims 1 to 7, characterized in that The information of the first data set is the statistical information of the first data set, and the statistical distance is the distance between the statistical information of the second data set of each of the N terminal devices and the statistical information of the first data set.

10. The method according to claim 8 or 9, characterized in that, The distance includes at least one of the following: Fréchet distance; maximum mean discrepancy; earth mover's (Wasserstein) distance; Kolmogorov-Smirnov distance.

11. A node selection method, characterized in that, including: Receiving first information from a network device, the first information including information of a first data set of the network device; Sending second information, the second information including a statistical distance, the statistical distance being used to indicate the distance between a second data set of a terminal device and the first data set, the second information being used to determine M terminal devices, where M is a positive integer; Receiving indication information, the indication information being used to indicate the M terminal devices; In response to the indication information, sending the second data set, the second data set being used for model training.

12. The method according to claim 11, characterized in that, The receiving the indication information includes; Receiving a broadcast signal, the broadcast signal including first indication information, the first indication information being used to indicate a set of the M terminal devices; or, Receiving second indication information, the second indication information being used to indicate the terminal device.

13. The method according to claim 11 or 12, characterized in that, The second information further includes communication capabilities, the communication capabilities including at least one of the following: the distance between the terminal device and the network device, the energy consumption per unit data volume of the terminal device.

14. The method according to any one of claims 11 to 13, characterized in that The information of the first data set is the first data set, and the statistical distance is the distance between the second data set and the first data set.

15. The method according to any one of claims 11 to 13, characterized in that The information of the first data set is the statistical information of the first data set, and the statistical distance is the distance between the statistical information of the second data set and the statistical information of the first data set.

16. The method according to claim 14 or 15, characterized in that, The distance includes at least one of the following: Fréchet distance; maximum mean discrepancy; earth mover's (Wasserstein) distance; Kolmogorov-Smirnov distance.

17. A communication device, characterized in that, including a module or unit for performing the method according to any one of claims 1 to 10.

18. A communication device, characterized in that, including: A processor and a memory, the processor being coupled to the memory and configured to read and execute instructions in the memory to perform the method according to any one of claims 1 to 10.

19. A communication device, characterized in that, including a module or unit for performing the method according to any one of claims 11 to 16.

20. A communication device, characterized in that, including: A processor and a communication interface, the communication interface being connected to the processor, the communication interface being configured to obtain a program or instructions, and the processor being configured to execute the method according to any one of claims 1 to 10 or execute the method according to any one of claims 11 to 16 by running the program or instructions.

21. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program code that, when run on a computer, causes the computer to execute the method according to any one of claims 1 to 10 or claims 11 to 16.

22. A chip or a chip system, characterized in that, Comprising: a circuit configured to execute the method according to any one of claims 1 to 10 or claims 11 to 16.

23. A communication system, characterized in that, Comprising: A first device configured to execute the method according to any one of claims 1 to 10; A second device configured to execute the method according to any one of claims 11 to 16.

Images