Electronic device for federated learning server and electronic device for core network

Abstract

The invention relates to an electronic device for a federated learning server and an electronic device for a core network. The electronic device for the federated learning server comprises at least one processor and at least one memory, the at least one memory comprises computer program code, and the at least one memory and the computer program code are configured to be connected with the at least one processor through the at least one processor. The method comprises the following steps: sending magnitude information about a parameter magnitude of a global model for federated learning to a core network to assist the core network in identifying federated learning clients which cannot upload all parameters of a local model from a plurality of federated learning clients as partial uploading clients, and under the condition that a preset condition is met, carrying out aggregation based on partial parameters of the local model received from the partial uploading client, so as to obtain a global model in the current round of training of federated learning.

Description

Technical Field

[0001] The present disclosure relates to the field of wireless communication technologies, and more particularly to an electronic device for a federated learning server, an electronic device for a core network, and an electronic device for a federated learning client. More specifically, it relates to an electronic device for a federated learning server, an electronic device for a core network, and an electronic device for a federated learning client for federated learning based on partial parameters of a local model. Background Art

[0002] With the continuous growth of terminal intelligent services, the requirements for wireless networks to support intelligent models are getting higher and higher. Generally, the data sets of user equipment (UE) are different, and UEs are reluctant to share their data sets due to issues such as information privacy, resulting in data barrier problems. In addition, the unique data sets of UEs are often small in size and difficult to support the training of more accurate and generalizable machine learning (ML) models. Based on this, federated learning (FL) breaks the data barrier problem among UEs through a training mode in which an FL client (e.g., a UE) locally trains a local model, an FL server aggregates the local models of multiple FL clients into a global model, and distributes the global model to the FL clients, and can further train an ML model with high accuracy and strong generalization. Summary of the Invention

[0003] A brief overview of the present invention is given below to provide a basic understanding of certain aspects of the present invention. It should be understood that this overview is not an exhaustive overview of the present invention. It is not intended to identify the key or important parts of the present invention, nor is it intended to limit the scope of the present invention. Its purpose is merely to present certain concepts in a simplified form as a prelude to a more detailed description to be discussed later.

[0004] According to one aspect of the present disclosure, there is provided an electronic device for a federated learning server, the electronic device including at least one processor and at least one memory, the at least one memory including computer program code, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform: sending magnitude information about the number of parameters of a global model for federated learning to a core network to assist the core network in identifying, from among multiple federated learning clients, a federated learning client that cannot upload all the parameters of its local model as a partial upload client, and, when a predetermined condition is satisfied, aggregating based on partial parameters of the local model received from the partial upload client to obtain a global model in the current round of training of federated learning.

[0005] According to one aspect of the present disclosure, there is provided an electronic device for a core network, the electronic device including at least one processor and at least one memory, the at least one memory including computer program code, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform: receiving, from a federated learning server, magnitude information about the parameter magnitude of a global model for federated learning, to identify, from among a plurality of federated learning clients, a federated learning client that cannot upload all parameters of its local model as a partial upload client, for, when a predetermined condition is satisfied, the federated learning server to aggregate based on partial parameters of the local model received from the partial upload client to obtain a global model in the current round of training of the federated learning.

[0006] According to one aspect of the present disclosure, there is provided an electronic device for a federated learning client, the electronic device including at least one processor and at least one memory, the at least one memory including computer program code, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform: when a predetermined condition is satisfied, sending, to a federated learning server, partial parameters of a local model for federated learning, for the federated learning server to aggregate the local model based on the partial parameters to obtain a global model in the current round of training of the federated learning, wherein the federated learning server sends magnitude information about the parameter magnitude of the global model to the core network to assist the core network in identifying the electronic device as a partial upload client that cannot upload all parameters of its local model.

[0007] According to one aspect of the present disclosure, there is provided a method for a federated learning server, including: sending, to a core network, magnitude information about the parameter magnitude of a global model for federated learning, to assist the core network in identifying, from among a plurality of federated learning clients, a federated learning client that cannot upload all parameters of its local model as a partial upload client, and when a predetermined condition is satisfied, aggregating based on partial parameters of the local model received from the partial upload client to obtain a global model in the current round of training of the federated learning.

[0008] According to one aspect of the present disclosure, there is provided a method for a core network, including: receiving, from a federated learning server, magnitude information about the parameter magnitude of a global model for federated learning, to identify, from among a plurality of federated learning clients, a federated learning client that cannot upload all parameters of its local model as a partial upload client, for, when a predetermined condition is satisfied, the federated learning server to aggregate based on partial parameters of the local model received from the partial upload client to obtain a global model in the current round of training of the federated learning.

[0009] According to one aspect of the present disclosure, a method for a federated learning client is provided, including: when a predetermined condition is met, sending partial parameters of a local model for federated learning to a federated learning server, so that the federated learning server aggregates the local model based on the partial parameters to obtain a global model in the current round of training of the federated learning, wherein the federated learning server sends magnitude information about the parameter magnitude of the global model to a core network to assist the core network in identifying the federated learning client as a partial upload client that cannot upload all parameters of its local model.

[0010] According to other aspects of the present invention, computer program code and computer program products for implementing the above method, and a computer-readable storage medium having recorded thereon the computer program code for implementing the above method are also provided. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] To further elaborate the above and other advantages and features of the present invention, the following detailed description of specific embodiments of the present invention is provided in conjunction with the accompanying drawings. The accompanying drawings are included in and form a part of this specification together with the following detailed description. Elements having the same function and structure are denoted by the same reference numerals. It should be understood that these drawings only depict typical examples of the present invention and should not be regarded as limiting the scope of the present invention. In the drawings:

[0012] Figure 1 A schematic diagram of federated learning in a vehicle network scenario in the prior art is shown;

[0013] Figure 2 A functional module block diagram of an electronic device for a federated learning server according to an embodiment of the present disclosure is shown;

[0014] Figure 3 A diagram showing an example of a neural network model;

[0015] Figure 4 A schematic diagram showing clustering using the K-Means algorithm according to an embodiment of the present disclosure;

[0016] Figure 5 A schematic diagram showing partial parameter upload according to an embodiment of the present disclosure;

[0017] Figure 6 A schematic diagram of a lightweight federated learning model training process according to an embodiment of the present disclosure;

[0018] Figure 7 Another schematic diagram of a lightweight federated learning model training process according to an embodiment of the present disclosure;

[0019] Figure 8Shows a functional module block diagram of an electronic device for a core network according to another embodiment of the present disclosure;

[0020] Figure 9 Shows a functional module block diagram of an electronic device for a federated learning client according to another embodiment of the present disclosure;

[0021] Figure 10 Shows a flowchart of a method for a federated learning server according to an embodiment of the present disclosure;

[0022] Figure 11 Shows a flowchart of a method for a core network according to another embodiment of the present disclosure;

[0023] Figure 12 Shows a flowchart of a method for a federated learning client according to yet another embodiment of the present disclosure;

[0024] Figure 13 Is a block diagram showing a first example of a schematic configuration of an eNB or gNB to which the technology of the present disclosure can be applied;

[0025] Figure 14 Is a block diagram showing a second example of a schematic configuration of an eNB or gNB to which the technology of the present disclosure can be applied;

[0026] Figure 15 Is a block diagram showing an example of a schematic configuration of a smart phone to which the technology of the present disclosure can be applied;

[0027] Figure 16 Is a block diagram showing an example of a schematic configuration of an in-vehicle navigation device to which the technology of the present disclosure can be applied; and

[0028] Figure 17 Is a block diagram of an exemplary structure of a general-purpose personal computer in which the method and / or apparatus and / or system according to an embodiment of the present invention can be implemented. Detailed Description of the Invention

[0029] Hereinafter, exemplary embodiments of the present invention will be described with reference to the accompanying drawings. For clarity and conciseness, not all features of the actual embodiments are described in the specification. However, it should be understood that many implementation-specific decisions must be made in the process of developing any such actual embodiment in order to achieve the specific goals of the developer, for example, to comply with those limitations related to the system and business, and these limitations may vary with different embodiments. In addition, it should be understood that although the development work may be very complex and time-consuming, for those skilled in the art who benefit from the present disclosure, such development work is merely a routine task.

[0030] Here, it should also be noted that in order to avoid obscuring the present invention with unnecessary details, only the device structures and / or processing steps closely related to the solution according to the present invention are shown in the drawings, while other details less related to the present invention are omitted.

[0031] Figure 1 The figure shows a schematic diagram of federated learning in a vehicle networking scenario in the prior art.

[0032] As Figure 1 shown, during the FL training process, within a certain communication range, multiple FL clients (for example, Figure 1 the shown FL client 1 to FL client 7) locally train local models and upload the local models to the FL server. The FL server aggregates all the local models and obtains the global FL model to complete the FL model training.

[0033] In this application, the vehicle networking scenario is taken as an example for description, but this application is not limited to vehicle networking. This application can be applied to all scenarios using FL for model training. In the following description, for example, in the vehicle networking scenario, the FL server is deployed on the base station and has AF (Application Function), and the FL client is a mobile vehicle (UE).

[0034] Since the FL model parameters are huge, uploading all the parameters of the local model by the FL client will increase the communication overhead. Therefore, this application proposes a lightweight FL model training in which the FL client only uploads part of the parameters of the local model. The lightweight FL model training is initiated by the FL server.

[0035] According to an embodiment of the present disclosure, there is provided an electronic device 2000 for a federated learning server. The electronic device 2000 includes at least one processor and at least one memory. The at least one memory includes computer program code, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device 2000 to execute: sending magnitude information about the number of parameters of the global model for federated learning to the core network to assist the core network in identifying, from multiple federated learning clients, the federated learning clients that cannot upload all the parameters of their local models as partial upload clients, and aggregating, under the condition of meeting a predetermined condition, based on the partial parameters of the local models received from the partial upload clients to obtain the global model in the current round of training of the federated learning.

[0036] Figure 2 The figure shows a functional module block diagram of an electronic device 2000 for a federated learning server according to an embodiment of the present disclosure.

[0037] As Figure 2As shown, the electronic device 2000 serving as an FL server includes: a server control unit 2001 that performs control; a processing unit 2003 that can be configured to send, under the control of the server control unit 2001, magnitude information regarding the parameter magnitude of the global model for federated learning to a core network to assist the core network in identifying, from among multiple federated learning clients, a federated learning client that cannot upload all the parameters of its local model as a partial upload client; and an aggregation unit 2005 that can be configured to, under the control of the server control unit 2001 and when a predetermined condition is met, perform aggregation based on the partial parameters of the local models received from the partial upload clients to obtain the global model in the current round of training of federated learning.

[0038] Among them, the server control unit 2001 can be implemented as one or more processing circuits and at least one memory. The processing circuit can be implemented as a processor or a chip, etc., and the at least one memory can be RAM, ROM, etc. The at least one memory is used to store, for example, computer program code and data required for the processing circuit to execute processing. The processing unit 2003 and the aggregation unit 2005 can perform operations under the control of the server control unit 2001. And it should be understood that Figure 2 each functional unit in the electronic device 2000 shown is only a logical module divided according to the specific functions it implements, rather than for limiting the specific implementation manner.

[0039] The electronic device 2000 can be set, for example, on the base station side or communicatively connected to a base station. For example, the electronic device 2000 can operate as the base station itself and can also include external devices such as a memory, a transceiver (not shown), etc. The memory can be used to store programs and related data information required for the electronic device 2000 to implement various functions. The transceiver can include one or more communication interfaces to support communication with different devices (such as UEs, base stations, etc.). The implementation form of the transceiver is not specifically limited here.

[0040] As an example, the base station can be an eNB or a gNB, for example.

[0041] As an example, the core network can be a Network Data Analytics Function (NWDAF) network element, for example.

[0042] The wireless communication system according to the present disclosure may be a 5G NR (New Radio) communication system, a 5G+ communication system, or a 6G communication system. Further, the wireless communication system according to the present disclosure may include a Non-terrestrial network (NTN). Optionally, the wireless communication system according to the present disclosure may further include a Terrestrial network (TN). Additionally, those skilled in the art can understand that the wireless communication system according to the present disclosure may also be a 4G or 3G communication system.

[0043] For example, some upload clients are FL clients that cannot upload all the parameters of their local models due to poor channel quality.

[0044] For example, the above-mentioned multiple federated learning clients are all the clients participating in federated learning.

[0045] Combined with Figure 1 As can be seen from the federated learning in the vehicle network scenario in the prior art shown in the figure, if the FL client uploads all its model parameters in each training iteration, it will cause a large amount of communication overhead. Additionally, in some cases where some FL clients cannot upload all the parameters of their local models, forcing these FL clients to upload all the parameters of their local models makes it easy for transmission errors to occur during the upload of the local models, resulting in errors in the local models received by the FL server, thereby damaging the accuracy of the global model.

[0046] The electronic device 2000 according to an embodiment of the present disclosure sends magnitude information about the parameter magnitude of the global model to the core network to assist the core network in identifying some upload clients, so that during the upload of the local model by some upload clients, only some model parameters need to be uploaded, thereby reducing the communication overhead of FL model training and not affecting the accuracy of the global model.

[0047] Hereinafter, the electronic device 2000 is sometimes referred to as the FL server.

[0048] For example, under the condition of meeting the above-mentioned predetermined conditions, the electronic device 2000 aggregates the partial parameters of the local model uploaded by some upload clients and the all parameters of the local model uploaded by non-some upload clients to obtain the global model in the current round of training of federated learning.

[0049] For example, under the condition of not meeting the predetermined conditions, the above-mentioned multiple federated learning clients upload all the parameters of the local model to the electronic device 2000, and the electronic device 2000 aggregates based on all the parameters of the received local model to obtain the global model in the current round of training of federated learning.

[0050] As an example, the magnitude information (hereinafter, sometimes referred to as the model magnitude) includes information about the size of the parameters of the global model (which can be simply referred to as the model size). Those skilled in the art can also think of other examples of magnitude information, which will not be elaborated here.

[0051] For example, the untrusted FL server sends the magnitude information to the NWDAF network element through the Network Exposure Function (NEF) network element. The trusted FL server directly sends the magnitude information to the NWDAF network element.

[0052] As an example, the magnitude information is used as an input for UE communication analysis in the NWDAF, and the UE communication prediction output of the UE communication analysis includes a list with the IDs of the partial upload clients.

[0053] For UE Communication Analytics in the NWDAF, see section 6.7.3 of 3GPP TS23.288. For example, the magnitude information can be placed in the input of the service, and the list of FL clients that cannot upload all the model parameters can be placed in the UE Communication Predictions output of the service.

[0054] For example, the newly added input parameters can be as follows:

[0055]

[0056] For example, the newly added output parameters can be as follows:

[0057]

[0058] As an example, the predetermined conditions include receiving a list of partial upload clients from the core network and / or the accuracy of the global model in the previous round of training reaching a preset accuracy. Among them, the list includes the IDs of the partial upload clients, and in the case where the accuracy of the global model in the previous round of training reaches the preset accuracy, multiple federated learning clients all act as partial upload clients.

[0059] For example, after the NWDAF network element determines the partial upload clients by identifying that some FL clients cannot upload all the parameters of their local models due to poor channel quality, it stores the IDs of the partial upload clients in a list and sends the list to the FL server to assist the FL server in subsequently notifying these partial upload clients to only upload partial parameters.

[0060] For example, after the global model reaches the preset accuracy, it is no longer necessary to upload all the parameters of the local model, and only the model parameters that are currently most in need of update can be uploaded. As an example, those skilled in the art can preset the preset accuracy according to experience or application scenarios. For example, the preset accuracy can be 95%, the preset accuracy can be 80%, etc., which will not be elaborated here.

[0061] For example, in the case where the accuracy of the global model in the previous round of training reaches the preset accuracy, all multiple federated learning clients are considered as partial upload clients. For example, in the case where the list is empty and the accuracy of the global model in the previous round of training reaches the preset accuracy, although there is no federated learning client that cannot upload all the parameters of its local model, all multiple federated learning clients are considered as partial upload clients. In addition, in the case where the accuracy of the global model in the previous round of training reaches the preset accuracy, the method for determining partial parameters is the same as the method for determining partial parameters in the case of an FL client that cannot upload all the parameters of its local model due to poor channel quality. The method for determining partial parameters will be described in detail below.

[0062] Combined with Figure 1 As can be seen from the federated learning in the vehicle network scenario in the prior art shown in Figure 1 , since the model parameters are huge, the communication overhead will increase when the client uploads all the parameters of the local model. In particular, when the channel quality of some FL clients is poor and / or the global model has reached the preset accuracy, uploading all the parameters of the local model by the FL client will further increase the communication overhead. In addition, in the case where the channel quality of some FL clients is poor, forcing the upload of all the parameters of the local models of these FL clients makes it easy to generate transmission errors during the upload of the local models, and there are errors in the local models received by the FL server, thereby damaging the accuracy of the global model. For example, in

[0063] In contrast, in the embodiments according to the present disclosure, in the case where the FL server receives a list including FL clients that cannot upload all the parameters of the model, the FL server triggers lightweight FL model training (i.e., partial upload clients only upload partial parameters of the local model), which can not only reduce the communication overhead, but also reduce the error of the local models received by the FL server, thus not affecting the accuracy of the global model. When the accuracy of the global model of the FL server reaches the preset preset accuracy (model accuracy threshold), performing lightweight FL model training can not only reduce the communication overhead, but also does not affect the accuracy of the global model.

[0064] In the case where the accuracy of the global model reaches the preset accuracy, since the accuracy difference in each iteration of FL training is not particularly large, lightweight FL model training can be periodically triggered, that is, some parameters of the local model are uploaded; in the case where the communication quality of the FL client is poor, since each FL training iteration requires the client to upload its local model, it is necessary to predict the communication quality of the FL client in each round of FL training iteration to determine whether the FL client can upload all the parameters of its local model. And for example, data transmission between the 5G core network (5GC) and the FL server without occupying wireless communication resources does not cause a large communication overhead. Therefore, in this case, lightweight FL model training can be triggered in real time.

[0065] When the list of FL clients returned by the NWDAF network element is empty and the accuracy of the current global model does not reach the preset accuracy, the FL server sends a notification to the FL client to upload all the parameters of the local model.

[0066] As an example, the IDs of some of the upload clients included in the above list are determined by the core network based on the magnitude information and the channel information between the electronic device 2000 and the federated learning client obtained by prediction.

[0067] For example, the NWDAF network element has the function of predicting the channel information between the FL server and the FL client (for example, the base station and the UE). The NWDAF network element uses the above predicted information and the magnitude information of the received model parameters to determine whether the FL client can upload all the parameters of its local model in the current communication situation.

[0068] As an example, the channel information includes at least one of the channel quality indicator (CQI), precoding matrix indicator (PMI), reference signal received power (RSRP), and signal-to-noise ratio (SNR) of the channel between the electronic device 2000 and the federated learning client.

[0069] For example, the NWDAF network element calculates the transmission rate through the SNR information and bandwidth information between the FL server and the FL client, and calculates the amount of data that can be uploaded within the above-mentioned delay limit based on the FL local model upload transmission delay limit, and then determines whether the FL client can upload all the parameters of its local model.

[0070]

[0071] In expression (1), SNR is the SNR value of the channel between the FL server and the FL client, B is the bandwidth allocated by the FL server for the FL client, Delay is the upper limit of the FL local model upload transmission delay restricted by the FL server, is the amount of data that the FL client can support for uploading, The data volume of the parameters of the FL global model. By comparing with it is possible to determine whether the FL client can upload all the parameters of the local model, and how much proportion of the model parameters can be supported for upload in the case where all the parameters of the local model cannot be uploaded. As can be seen from expression (1), when is greater than , it is determined that the FL client can upload all the parameters of its local model to the FL server under the current communication condition; while when is less than , it is determined that the FL client cannot upload all the parameters of its local model to the FL server under the current communication condition.

[0072] As an example, the processing unit 2003 can be configured to calculate, for the parameters of the global model in the previous round of training and the parameters of the local models uploaded by multiple federated learning clients in the previous round of training: the deviation degree of the local model parameters representing the dispersion degree among the parameters of the local models of multiple federated learning clients, and the deviation degree of the global model parameters representing the deviation degree between the parameters of the local models of multiple federated learning clients and the parameters of the global model, and determine partial parameters based on the deviation degree of the local model parameters and the deviation degree of the global model parameters. That is, the electronic device 2000 can determine the parameters to be supplemented in the global model in the current round of training according to the local model parameters of the FL client received in the previous round of iteration (i.e., determine the partial parameters to be uploaded by some uploading clients in the current round of training).

[0073] As an example, the processing unit 2003 can be configured to calculate the average value of the i-th parameter based on the i-th parameter of the local models of multiple federated learning clients, and calculate the deviation degree of the local model parameters for the i-th parameter based on the difference between the i-th parameter of the local model of each federated learning client and the average value; and calculate the deviation degree of the global model parameters for the i-th parameter based on the difference between the i-th parameter of the local model of each federated learning client and the i-th parameter of the global model.

[0074] For example, the FL server calculates the deviation degree of the local model parameters for the i-th parameter according to the i-th parameter of the local model uploaded by the FL client in the previous round of iteration:

[0075]

[0076] In expression (2), N represents the number of multiple FL clients, represents the i-th parameter of the n-th FL client (n is a positive integer greater than or equal to 1 and less than or equal to N), Represents the average value of the i-th parameter among all N FL client local models.

[0077] As described above, σ l Measures the degree of dispersion of the local model parameters of multiple FL clients. For example, the greater the degree of dispersion of all FL clients for the i-th parameter, the greater the need for further update of this parameter, that is, the greater the probability that this parameter of the local model is uploaded to the FL server.

[0078] For example, the FL server calculates the global model parameter deviation degree for the i-th parameter based on the difference between the i-th parameter of the local model uploaded by the FL clients in the previous round of iteration and the i-th parameter p g (i) of the global model:

[0079]

[0080] As described above, σ g Measures the average gap between the local model parameters of all FL clients and the global model parameters. For example, for the i-th parameter in the model, if the average gap between the i-th parameter in all FL client local models and the i-th parameter in the global model is greater, the greater the need for further update of this parameter, that is, the greater the probability that this parameter is uploaded to the FL server.

[0081] If there are FL clients among the N FL clients that did not upload the i-th parameter in the previous round of training, then the of the FL clients that did not upload the i-th parameter is 0.

[0082] As an example, the processing unit 2003 can be configured to cluster the parameters using a clustering method based on the local model parameter deviation degree and the global model parameter deviation degree, so as to obtain multiple clusters after clustering, select the selected clusters that meet the predetermined cluster conditions from the multiple clusters, and use the parameters in the selected clusters as partial parameters.

[0083] For example, the global model and the local model are hierarchical neural network models. Generally, for a neural network model, the parameters of the neural network refer to weight values and bias values.

[0084] Figure 3 Is a diagram showing an example of a neural network model. As Figure 3 shown, the circles in the first column represent the neurons of the input layer, the circles in the last column represent the neurons of the output layer, the circles between the first column and the last column represent the neurons of the hidden layer, the connections between the neurons represent the weight values, and each neuron has a bias value. Taking Figure 3Take the first two layers of neurons as an example. The index of each neuron is the value in its circle (the index of neurons can also be defined in other ways as long as each neuron can be uniquely identified). For example, the weight value between neuron 1 and neuron 6 is named as ω 1，6 , and the weight value between neuron 5 and neuron 11 is named as ω 5，11 . The weight values between different pairs of neurons are named in this way by analogy. Take the neuron with index 6 as an example. The value of the neuron is calculated as shown in the following expression (4):

[0085] x6 = (ω 1，6 x1 + ω 2，6 x2 + ω 3，6 x3 + ω 4，6 x4 + ω 5，6 x5) + b6 Expression (4)

[0086] In expression (4), ω is the weight value and b is the bias value.

[0087] As can be seen from the above description, the number of weights and biases in the neural network model is huge. If the calculation and clustering of the local model parameter deviation degree and the global model parameter deviation degree of the weights and biases are carried out, the computational amount is large. Therefore, in this application, the global model and all uploaded local models are processed with all-1 inputs at the FL server side. Specifically, all the input values are set to 1, for example, [1, 1, 1, 1,...], and these values are input into the global model and all uploaded local models to obtain the value of each neuron in the model (the calculation process of the neuron value refers to the above expression (4) for neuron 6). Therefore, the parameters of the model described in this application refer to the values of the neurons in the FL model.

[0088] As an example, the predetermined cluster condition includes that the selected probability corresponding to the cluster is greater than the predetermined probability, and the processing unit 2003 can be configured to use the local model parameter deviation degree and the global model parameter deviation degree as the coordinates of a two-dimensional coordinate system respectively, and use the calculated local model parameter deviation degree and global model parameter deviation degree corresponding to the parameters as the coordinate values, so as to represent each parameter as a point in the coordinate system. The K-Means algorithm is used to cluster all the points in the coordinate system to obtain multiple clusters, and for each of at least some of the multiple clusters, based on the distance between the cluster center of the cluster and the origin of the coordinate system, and the included angle between the line connecting the cluster center to the origin of the coordinate system and the dividing line of the first quadrant of the coordinate system, calculate the selected probability of the cluster, where the dividing line is a straight line passing through the origin of the coordinate system and making a predetermined angle with the horizontal axis of the first quadrant, where the greater the distance, the greater the selected probability, and the smaller the included angle, the greater the selected probability.

[0089] For example, those skilled in the art can pre-determine a predetermined probability and a predetermined angle based on experience or application scenarios.

[0090] The samples in the K-Means algorithm are defined as parameters represented by the degree of deviation of local model parameters and the degree of deviation of global model parameters (i.e., points in the above coordinate system). The specific operation steps for clustering all points in the above coordinate system using the K-Means algorithm are as follows:

[0091] Step a: Randomly select K points from the samples as clustering centers (cluster centroids);

[0092] Step b: Calculate the distances between the other samples in the samples and the K clustering centers respectively, and assign these samples to the category of the clustering center with the closest distance;

[0093] Step c: Calculate the average value for each category of the clustered samples above, and solve for the new cluster centroids;

[0094] Step d: Compare with the K cluster centroids obtained in the previous calculation. If the cluster centroids change, go to step b; otherwise, go to step e;

[0095] Step e: When the centroids no longer change, stop the iteration and output the clustering result.

[0096] Step f: Select clusters from the clustering result, and use the parameters in the selected clusters as partial parameters.

[0097] Figure 4 is a schematic diagram showing clustering using the K-Means algorithm according to an embodiment of the present disclosure.

[0098] In Figure 4 , for example, a coordinate system is established with the degree of deviation of local model parameters σ l as the abscissa and the degree of deviation of global model parameters σ g as the ordinate. In view of the fact that all parameters in the FL model have σ l and σ g values, each parameter in the FL model is a point in the established coordinate system. All parameters in the FL are represented in the established coordinate system, and clustering / clustering of model parameters is achieved by means of the K-Means algorithm. The clustering result is schematically shown by the circles in Figure 4 (illustrative rather than limiting, Figure 4 shows that the clustering result includes cluster 1, cluster 2, and cluster 3), and each cluster has a cluster center.

[0099] In the following description, for simplicity, it is assumed that the predetermined angle is 45° for description. Then the above dividing line is a straight line at an angle of 45 degrees with the horizontal axis in the first quadrant.

[0100] For example, the expression for calculating the probability of a cluster being selected is as follows:

[0101]

[0102] In expression (5), K represents the total number of clusters in the clustering result (in Figure 4 , for example, K = 3), k represents the k-th cluster (in Figure 4 , for example, k ranges from 1 to 3), P(k) represents the probability of selecting the k-th cluster, d(k) represents the distance from the centroid of the k-th cluster to the coordinate origin o (in 4, d(1), d(2), and d(3) are shown), θ(k) represents the deviation angle between the line connecting the centroid of the k-th cluster and the coordinate origin and the x-axis (in 4, θ(3) is shown), and g(·) represents the normalization function.

[0103] In an embodiment according to the present disclosure, there are two principles for cluster selection. One is to select the cluster with a centroid far from the coordinate origin (ensuring a large combined value of σ l and σ g ), so d(k) is considered in expression (5). The other is to select the cluster with a small angle between the line connecting the centroid to the coordinate origin and the dividing line of the first quadrant of the coordinate system (for example, a line at a 45-degree angle to the x-axis, Figure 4 shown as the dashed line in l and σ g are both large, and there will be no extreme situation of bias towards one side, for example, avoiding the centroid being on the x-axis and far from the coordinate origin), so |θ(k) - 45°| is considered in the cluster selection expression (5). By calculating the P(k) values of all clusters, clusters are selected based on the P(k) values. For example, all the parameters within the cluster with the largest P(k) value can be selected as the partial parameters to be uploaded, that is, the cluster of parameters with the largest deviation degree (the deviation degree of local model parameters σ l and the deviation degree of global model parameters σ g are both as large as possible) is selected as the partial parameters to be filled.

[0104] In the above, clustering is described by taking the K-Means algorithm as an example. However, those skilled in the art can think of other clustering methods, which will not be elaborated here.

[0105] As an example, the processing unit 2003 can be configured to send notification information about the partial parameters to the partial upload client, so that the partial upload client only uploads the partial parameters in the current round of training.

[0106] As described above, the global model is a hierarchical neural network model. As an example, the notification information includes the position information of the partial parameters in the neural network model. This position information is, for example, the index of the neuron in the above.

[0107] As an example, the aggregation unit 2005 can be configured to align partial parameters of a local model received from a partial upload client and all parameters of local models received from other clients among multiple federated learning clients based on location information, and perform aggregation of the local models.

[0108] Figure 5 It is a schematic diagram showing partial parameter upload according to an embodiment of the present disclosure.

[0109] After receiving the notification information of partial parameter upload (for example, a list including the indices of neurons to be uploaded), the FL client uploads partial parameters of the model. For example, Figure 5 The neurons to be uploaded (abbreviated as uploaded neurons) are shown in. The FL client uploads the weight values of all neurons in the upper layer connected to the uploaded neuron and the bias value of the uploaded neuron to the FL server. After receiving the local models uploaded by all FL clients, the FL server aligns them according to the indices of the neurons for aggregation to obtain the global model.

[0110] Figure 6 It is a schematic diagram of the lightweight FL model training process according to an embodiment of the present disclosure. In Figure 6 it is assumed that the FL server (electronic device 2000) is not trusted, and the multiple federated learning clients mentioned above include FL client 1 to FL client N.

[0111] In step 1, the FL server performs FL client selection. For example, the FL server selects the FL clients that need to participate in FL training according to requirements. For example, since the value of the dataset owned by each FL client is different (that is, the datasets owned by each FL client are heterogeneous), the FL server selects FL clients based on the value of the dataset to fully train the FL model.

[0112] In step 2, the FL server performs initial global model distribution. After selecting the FL clients, the FL server distributes the initial global model to the selected FL clients for subsequent local model training of the FL clients.

[0113] In step 3 (including steps 3a and 3b), the FL server performs model parameter magnitude notification. As shown in steps 3a and 3b, the FL server sends model parameter magnitude notification to the NWDAF network element through the NEF network element to assist the subsequent NWDAF network element in determining which FL clients cannot upload all parameters of their local models under the current communication conditions.

[0114] In step 4, the NWDAF network element conducts a comparative analysis of the model parameter order of magnitude and the channel quality of the FL clients to determine whether the FL clients can upload all the parameters of their local models to the FL server under the current communication conditions.

[0115] In step 5 (including steps 5a and 5b), the NWDAF network element sends the list of FL clients that cannot upload all the model parameters to the FL server.

[0116] In step 6, the FL server determines the supplementation of partial model parameters. Under the condition of meeting the predetermined conditions mentioned above, the FL server determines the partial parameters of the local model that need to be uploaded.

[0117] In step 7, the FL server notifies the upload of the local model (all model parameters or partial model parameters). For example, after the FL server determines the model parameters that need to be uploaded (supplemented), it notifies the FL clients which parameters should be uploaded in the current iteration round through a notification message including the position information of the model parameters to be supplemented. For example, the notification message is a list that stores the indices of the positions of all the model parameters that need to be supplemented. In addition, for the FL clients that are not notified of the supplementation of model parameters, they only need to upload all the parameters of their local models.

[0118] In step 8, the FL clients upload the local model (all model parameters or partial model parameters). For example, after the FL clients receive the notification of the upload of the local model (all parameters or partial parameters), in the case where the channel quality of some FL clients is poor, the FL clients with poor channels upload the corresponding model parameters in the local model according to the position information of the partial parameters to be uploaded, and other FL clients upload all the parameters of their local models; in the case where the global model accuracy reaches the preset model accuracy threshold, all FL clients upload the corresponding model parameters in the local model according to the position information of the partial parameters. In the above cases, when the FL clients upload the partial parameters of the local model, they upload the parameter value information at the corresponding positions according to the position information of the parameters.

[0119] In step 9, the FL server conducts model aggregation. For example, after the FL server receives the local models uploaded by all FL clients, the FL server aligns the local model parameters of the clients and completes the global model aggregation according to the position information and parameter value information of the local model parameters.

[0120] In step 10, the FL server updates / terminates the process. For example, according to the requirements of the aggregator task initiator, the FL server decides whether to continue the FL training process or terminate it.

[0121] In step 11, the FL server distributes the aggregated global model. The FL server distributes the globally aggregated model of this round of iteration to the FL clients for the next round of local model training by the FL clients.

[0122] In step 12, the FL clients update the local models. After receiving the new global model, the FL clients start a new round of training and update the local models.

[0123] Among them, in each round of training, steps 4 to 12 are iteratively executed.

[0124] Figure 7 is another schematic diagram of the lightweight FL model training process according to an embodiment of the present disclosure. In Figure 7 , it is assumed that the FL server (electronic device 2000) is trusted.

[0125] As described above, the untrusted FL server sends the model parameter magnitude notification to the NWDAF network element through the NEF network element, while the trusted FL server directly sends the model parameter magnitude notification to the NWDAF network element. Therefore, Figure 7 there is no NEF network element, and except for steps 3 and 5, Figure 7 the other steps of Figure 6 are the same as those in

[0126] The present disclosure also provides an electronic device 8000 for a core network according to another embodiment of the present disclosure. The electronic device 8000 includes at least one processor and at least one memory. The at least one memory includes computer program code. Among them, the at least one memory and the computer program code are configured to enable the electronic device 8000 to execute: receiving, from a federated learning server, magnitude information about the parameter magnitude of a global model for federated learning to identify, from multiple federated learning clients, a federated learning client that cannot upload all the parameters of its local model as a partial upload client, so that, when a predetermined condition is met, the federated learning server aggregates based on the partial parameters of the local model received from the partial upload clients to obtain a global model in the current round of training of the federated learning.

[0127] Figure 8 shows a functional module block diagram of an electronic device 8000 for a core network according to another embodiment of the present disclosure.

[0128] As Figure 8As shown, the electronic device 8000 includes: a core network control unit 8001 for control; an identification unit 8003 which, under the control of the core network control unit 8001, receives magnitude information regarding the parameter magnitude of the global model for federated learning from a federated learning server, so as to identify, from among multiple federated learning clients, the federated learning clients that cannot upload all the parameters of their local models as partial upload clients, such that, when a predetermined condition is met, the federated learning server aggregates based on the partial parameters of the local models received from the partial upload clients to obtain the global model in the current round of training of federated learning.

[0129] Among them, the core network control unit 8001 can be implemented as one or more processing circuits and at least one memory. The processing circuit can be implemented as a processor or a chip, etc. The at least one memory can be RAM, ROM, etc. The at least one memory is used to store computer program code and data required for the processing circuit to execute processing, etc. The identification unit 8003 can perform operations under the control of the core network control unit 8001. And it should be understood that Figure 8 each functional unit in the electronic device 8000 shown is only a logical module divided according to its specific implemented function, rather than for limiting the specific implementation manner.

[0130] For example, the electronic device 8000 can operate as the core network device itself and can also include external devices such as a memory, a transceiver (not shown), etc. The memory can be used to store programs and related data information required for the electronic device 8000 to implement various functions. The transceiver can include one or more communication interfaces to support communication with different devices (such as UEs, base stations, etc.). The implementation form of the transceiver is not specifically limited here.

[0131] As an example, the federated learning server in the embodiment of the electronic device 8000 can be the electronic device 2000 mentioned above. As an example, the electronic device 8000 can be the core network involved in the embodiment of the above-mentioned electronic device 2000.

[0132] For example, the partial upload clients are FL clients that cannot upload all the parameters of their local models due to poor channel quality.

[0133] For example, the above-mentioned multiple federated learning clients are all the clients participating in federated learning.

[0134] The electronic device 8000 according to an embodiment of the present disclosure can identify partial upload clients based on the magnitude information about the number of parameters of the global model received from the FL server, so that during the process of uploading the local model, the partial upload clients only need to upload partial model parameters, thereby reducing the communication overhead of FL model training and not affecting the global model accuracy.

[0135] As an example, the magnitude information is used as the input of UE communication analysis in NWDAF, and the UE communication prediction output of the UE communication analysis includes a list with the IDs of partial upload clients.

[0136] As an example, the magnitude information includes information about the size of the parameters of the global model. Those skilled in the art can also think of other examples of magnitude information, which will not be elaborated here.

[0137] For example, the electronic device 8000 can be an NWDAF network element. For example, an untrusted FL server sends the magnitude information to the NWDAF network element through the NEF network element. The trusted FL server directly sends the magnitude information to the NWDAF network element.

[0138] As an example, the predetermined condition includes that the federated learning server receives a list of partial upload clients from the electronic device 8000 and / or the accuracy of the global model in the previous round of training reaches a preset accuracy, where the list includes the IDs of partial upload clients.

[0139] For the description of the predetermined condition, please refer to the corresponding part in the embodiment of the electronic device 2000, which will not be elaborated here.

[0140] As an example, the identification unit 8003 can be configured to determine the IDs of the partial upload clients included in the list based on the magnitude information and the predicted channel information between the federated learning server and the federated learning client.

[0141] For example, the electronic device 8000 has the function of predicting the channel information between the FL server and the FL client (e.g., the base station and the UE). The electronic device 8000 determines whether the FL client can upload all the parameters of its local model to the FL server under the current communication situation based on the predicted information and the received magnitude information of the model parameters.

[0142] As an example, the channel information includes at least one of the channel quality indicator (CQI), precoding matrix indicator (PMI), reference signal received power (RSRP), and signal-to-noise ratio (SNR) of the channel between the electronic device 8000 and the federated learning client.

[0143] For the description of the electronic device 8000 determining the partial upload client based on the magnitude information and SNR, please refer to the description of Expression (1) in the embodiment of the electronic device 2000, which will not be repeated here.

[0144] The present disclosure also provides an electronic device 9000 for a federated learning client according to another embodiment of the present disclosure. The electronic device 9000 includes at least one processor and at least one memory. The at least one memory includes computer program code. Wherein, the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device 9000 to perform: when a predetermined condition is satisfied, send partial parameters of a local model for federated learning to a federated learning server, so that the federated learning server aggregates the local model based on the partial parameters to obtain a global model in the current round of training of the federated learning. Wherein, the federated learning server sends magnitude information about the number of parameters of the global model to a core network to assist the core network in identifying the electronic device 9000 as a partial upload client that cannot upload all parameters of its local model.

[0145] Figure 9 A functional module block diagram of an electronic device 9000 for a federated learning client according to another embodiment of the present disclosure is shown.

[0146] As Figure 9 shown, the electronic device 9000 includes: a client control unit 9001 for controlling; a communication unit 9003, which, under the control of the client control unit 9001, when a predetermined condition is satisfied, sends partial parameters of a local model for federated learning to a federated learning server, so that the federated learning server aggregates the local model based on the partial parameters to obtain a global model in the current round of training of the federated learning. Wherein, the federated learning server sends magnitude information about the number of parameters of the global model to a core network to assist the core network in identifying the electronic device 9000 as a partial upload client that cannot upload all parameters of its local model.

[0147] Wherein, the client control unit 9001 can be implemented as one or more processing circuits and at least one memory. The processing circuit can be implemented as a processor or a chip, etc. The at least one memory can be a RAM, a ROM, etc. The at least one memory is used to store computer program code and data required for the processing circuit to perform processing, etc. The communication unit 9003 can perform operations under the control of the client control unit 9001. And it should be understood that Figure 9 each functional unit in the electronic device 9000 shown in

[0148] For example, the electronic device 9000 can operate as a user equipment itself and can also include external devices such as a memory, a transceiver (not shown in the figure), etc. The memory can be used to store programs and related data information that the user equipment needs to execute to implement various functions. The transceiver can include one or more communication interfaces to support communication with different devices (e.g., base stations, other user equipments, etc.), and the implementation form of the transceiver is not specifically limited here.

[0149] As an example, the federated learning server in the embodiment of the electronic device 9000 can be the electronic device 2000 mentioned above, and the core network in the embodiment of the electronic device 9000 can be the electronic device 8000 mentioned above. As an example, the electronic device 9000 can be a partial upload client involved in the embodiments of the above-mentioned electronic device 2000 and electronic device 8000.

[0150] For example, the electronic device 9000 is an FL client that cannot upload all parameters of its local model due to poor channel quality.

[0151] During the process of uploading the local model, the electronic device 9000 according to the embodiment of the present disclosure only needs to upload some model parameters, thereby being able to reduce the communication overhead of FL model training and not affecting the global model accuracy.

[0152] As an example, the magnitude information is used as an input for UE communication analysis in NWDAF, and the UE communication prediction output of the UE communication analysis includes a list with the IDs of partial upload clients.

[0153] As an example, the magnitude information includes information about the size of the parameters of the global model. Those skilled in the art can also think of other examples of the magnitude information, which will not be elaborated here.

[0154] As an example, the predetermined conditions include that the federated learning server receives a list including the IDs of partial upload clients from the core network and / or the accuracy of the global model in the previous round of training reaches a preset accuracy.

[0155] For the description of the predetermined conditions, please refer to the corresponding part in the embodiment of the electronic device 2000, which will not be elaborated here.

[0156] As an example, the communication unit 9003 can be configured to receive notification information about partial parameters from the federated learning server.

[0157] As an example, the global model is a hierarchical neural network model, and the notification information includes the position information of the partial parameters in the neural network model.

[0158] For the description of the position information of some parameters in the neural network model, please refer to the corresponding part in the embodiment of the electronic device 2000, which will not be repeated here.

[0159] In the process of describing the electronic device 2000 and the electronic device 9000 in the above embodiments, some processes or methods are obviously also disclosed. In the following text, a summary of these methods is given without repeating some details already discussed above. However, it should be noted that although these methods are disclosed in the process of describing the above electronic devices, these methods do not necessarily use the described components or are not necessarily executed by those components. For example, the above embodiments of the electronic device can be implemented partially or completely using hardware and / or firmware, while the methods discussed below can be completely implemented by computer-executable programs, although these methods can also use the hardware and / or firmware of the electronic device.

[0160] Figure 10 The flowchart of a method S1000 for a federated learning server according to an embodiment of the present disclosure is shown. Method S1000 starts at step S1002. In step S1004, magnitude information about the number of parameters of the global model for federated learning is sent to the core network to assist the core network in identifying, from among multiple federated learning clients, the federated learning clients that cannot upload all the parameters of their local models as partial upload clients. In step S1006, when a predetermined condition is met, aggregation is performed based on the partial parameters of the local models received from the partial upload clients to obtain the global model in the current round of training of the federated learning. Method S1000 ends at step S1008.

[0161] This method can be executed, for example, by the electronic device 2000 described above. For the specific details, please refer to the description of the related processing of the above-mentioned electronic device 2000, which will not be repeated here.

[0162] Figure 11 The flowchart of a method S1100 for a core network according to another embodiment of the present disclosure is shown. Method S1100 starts at step S1102. In step S1104, magnitude information about the number of parameters of the global model for federated learning is received from the federated learning server to identify, from among multiple federated learning clients, the federated learning clients that cannot upload all the parameters of their local models as partial upload clients, so that when a predetermined condition is met, the federated learning server performs aggregation based on the partial parameters of the local models received from the partial upload clients to obtain the global model in the current round of training of the federated learning. Method S1100 ends at step S1106.

[0163] This method can be executed, for example, by the electronic device 8000 described above. For specific details, reference can be made to the description of the related processing of the electronic device 8000 above, which will not be repeated here.

[0164] Figure 12 FIG. shows a flowchart of a method S1200 for a federated learning client according to another embodiment of the present disclosure. The method S1200 starts at step S1202. In step S1204, when a predetermined condition is satisfied, partial parameters of a local model for federated learning are sent to a federated learning server for the federated learning server to aggregate the local model based on the partial parameters to obtain a global model in the current round of training of the federated learning, wherein the federated learning server sends magnitude information about the number of parameters of the global model to a core network to assist the core network in identifying the federated learning client as a partial upload client that cannot upload all parameters of its local model. The method S1200 ends at step S1206.

[0165] This method can be executed, for example, by the electronic device 9000 described above. For specific details, reference can be made to the description of the related processing of the electronic device 9000 above, which will not be repeated here.

[0166] The technology of the present disclosure can be applied to various products.

[0167] The electronic device 2000 can be disposed on the base station side or connected to the base station. The base station can be implemented as any type of evolved Node B (eNB) or gNB (5G base station). The eNB includes, for example, a macro eNB and a small eNB. The small eNB can be an eNB that covers a cell smaller than a macro cell, such as a pico eNB, a micro eNB, and a home (femto) eNB. A similar situation can also apply to the gNB. Instead, the base station can be implemented as any other type of base station, such as a NodeB and a base transceiver station (BTS). The base station can include: a main body configured to control wireless communication (also referred to as base station equipment); and one or more remote radio heads (RRHs) disposed in a place different from the main body. In addition, various types of electronic devices can act as a base station by temporarily or semi-persistently performing base station functions.

[0168] The electronic device 9000 can be implemented as various user devices. The user device can be implemented as a mobile terminal (such as a smart phone, a tablet personal computer (PC), a notebook PC, a portable game terminal, a portable / dongle-type mobile router, and a digital imaging device) or a vehicle-mounted terminal (such as an automotive navigation device). The user device can also be implemented as a terminal that performs machine-to-machine (M2M) communication (also referred to as a machine type communication (MTC) terminal). In addition, the user device can be a wireless communication module (such as an integrated circuit module including a single wafer) installed on each of the above terminals.

[0169] [Application Examples of Base Stations]

[0170] (First Application Example)

[0171] Figure 13 FIG. is a block diagram showing a first example of a schematic configuration of an eNB or a gNB to which the technology of the present disclosure can be applied. Note that in the following description, the eNB is taken as an example, but the same can also be applied to the gNB. The eNB 800 includes one or more antennas 810 and a base station device 820. The base station device 820 and each antenna 810 can be connected to each other via an RF cable.

[0172] Each of the antennas 810 includes a single or multiple antenna elements (such as multiple antenna elements included in a multiple-input multiple-output (MIMO) antenna), and is used for the base station device 820 to transmit and receive wireless signals. As Figure 13 shown, the eNB 800 can include multiple antennas 810. For example, the multiple antennas 810 can be compatible with multiple frequency bands used by the eNB 800. Although Figure 13 shows an example in which the eNB 800 includes multiple antennas 810, the eNB 800 can also include a single antenna 810.

[0173] The base station device 820 includes a controller 821, a memory 822, a network interface 823, and a wireless communication interface 825.

[0174] The controller 821 can be, for example, a CPU or a DSP, and operates various functions of the higher layers of the base station device 820. For example, the controller 821 generates data packets based on the data in the signals processed by the radio communication interface 825, and transmits the generated packets via the network interface 823. The controller 821 can bundle data from multiple baseband processors to generate bundled packets, and transmit the generated bundled packets. The controller 821 can have logical functions for performing controls such as radio resource control, radio bearer control, mobility management, admission control, and scheduling. The control can be performed in combination with a nearby eNB or a core network node. The memory 822 includes a RAM and a ROM, and stores programs executed by the controller 821 and various types of control data (such as a terminal list, transmission power data, and scheduling data).

[0175] The network interface 823 is a communication interface for connecting the base station device 820 to the core network 824. The controller 821 can communicate with a core network node or another eNB via the network interface 823. In this case, the eNB 800 and the core network node or other eNBs can be connected to each other through logical interfaces (such as the S1 interface and the X2 interface). The network interface 823 can also be a wired communication interface or a radio communication interface for a radio backhaul line. If the network interface 823 is a radio communication interface, compared with the frequency band used by the radio communication interface 825, the network interface 823 can use a higher frequency band for radio communication.

[0176] The radio communication interface 825 supports any cellular communication scheme (such as Long Term Evolution (LTE) and LTE-Advanced), and provides a radio connection to terminals located in the cell of the eNB 800 via the antenna 810. The radio communication interface 825 generally can include, for example, a baseband (BB) processor 826 and an RF circuit 827. The BB processor 826 can perform, for example, encoding / decoding, modulation / demodulation, and multiplexing / demultiplexing, and perform various types of signal processing of layers (such as layer 1, Medium Access Control (MAC), Radio Link Control (RLC), and Packet Data Convergence Protocol (PDCP)). Instead of the controller 821, the BB processor 826 can have a part or all of the above logical functions. The BB processor 826 can be a memory storing a communication control program, or a module including a processor configured to execute the program and related circuits. The update program can change the functions of the BB processor 826. The module can be a card or a blade inserted into a slot of the base station device 820. Alternatively, the module can also be a chip mounted on the card or the blade. At the same time, the RF circuit 827 can include, for example, mixers, filters, and amplifiers, and transmits and receives radio signals via the antenna 810.

[0177] As Figure 13As shown, the wireless communication interface 825 may include multiple BB processors 826. For example, the multiple BB processors 826 may be compatible with multiple frequency bands used by the eNB 800. As Figure 13 As shown, the wireless communication interface 825 may include multiple RF circuits 827. For example, the multiple RF circuits 827 may be compatible with multiple antenna elements. Although Figure 13 An example is shown in which the wireless communication interface 825 includes multiple BB processors 826 and multiple RF circuits 827, but the wireless communication interface 825 may also include a single BB processor 826 or a single RF circuit 827.

[0178] When the electronic device 2000 is implemented as Figure 13 the eNB 800 shown, its transceiver may be implemented by the wireless communication interface 825. At least a part of the functions may also be implemented by the controller 821. For example, the controller 821 can reduce the communication overhead of FL model training by executing the functions of the units in the electronic device 2000 without affecting the global model accuracy.

[0179] (Second application example)

[0180] Figure 14 FIG. is a block diagram showing a second example of a schematic configuration of an eNB or gNB to which the technology of the present disclosure can be applied. Note that, similarly, the following description uses the eNB as an example, but the same can be applied to the gNB. The eNB 830 includes one or more antennas 840, a base station device 850, and an RRH 860. The RRH 860 and each antenna 840 may be connected to each other via an RF cable. The base station device 850 and the RRH 860 may be connected to each other via a high-speed line such as an optical fiber cable.

[0181] Each of the antennas 840 includes a single or multiple antenna elements (such as multiple antenna elements included in a MIMO antenna) and is used to transmit and receive wireless signals for the RRH 860. As Figure 14 As shown, the eNB 830 may include multiple antennas 840. For example, the multiple antennas 840 may be compatible with multiple frequency bands used by the eNB 830. Although Figure 14 An example is shown in which the eNB 830 includes multiple antennas 840, but the eNB 830 may also include a single antenna 840.

[0182] The base station device 850 includes a controller 851, a memory 852, a network interface 853, a wireless communication interface 855, and a connection interface 857. The controller 851, the memory 852, and the network interface 853 are the same as the controller 821, the memory 822, and the network interface 823 described with reference to Figure 14 above.

[0183] The wireless communication interface 855 supports any cellular communication scheme (such as LTE and LTE-Advanced), and provides wireless communication to terminals located in the sector corresponding to the RRH 860 via the RRH 860 and the antenna 840. The wireless communication interface 855 generally may include, for example, a BB processor 856. Except that the BB processor 856 is connected to the RF circuit 864 of the RRH 860 via the connection interface 857, the BB processor 856 is the same as the BB processor 826 described with reference to Figure 14 As Figure 14 shown, the wireless communication interface 855 may include multiple BB processors 856. For example, multiple BB processors 856 may be compatible with multiple frequency bands used by the eNB 830. Although Figure 14 an example where the wireless communication interface 855 includes multiple BB processors 856 is shown, the wireless communication interface 855 may also include a single BB processor 856.

[0184] The connection interface 857 is an interface for connecting the base station device 850 (wireless communication interface 855) to the RRH 860. The connection interface 857 may also be a communication module for communication in the above-mentioned high-speed line for connecting the base station device 850 (wireless communication interface 855) to the RRH 860.

[0185] The RRH 860 includes a connection interface 861 and a wireless communication interface 863.

[0186] The connection interface 861 is an interface for connecting the RRH 860 (wireless communication interface 863) to the base station device 850. The connection interface 861 may also be a communication module for communication in the above-mentioned high-speed line.

[0187] The wireless communication interface 863 transmits and receives wireless signals via the antenna 840. The wireless communication interface 863 generally may include, for example, an RF circuit 864. The RF circuit 864 may include, for example, mixers, filters, and amplifiers, and transmits and receives wireless signals via the antenna 840. As Figure 14 shown, the wireless communication interface 863 may include multiple RF circuits 864. For example, multiple RF circuits 864 may support multiple antenna elements. Although Figure 14 an example where the wireless communication interface 863 includes multiple RF circuits 864 is shown, the wireless communication interface 863 may also include a single RF circuit 864.

[0188] The electronic device 2000 when implemented as Figure 14When the eNB 830 shown is considered, its transceiver can be implemented by the wireless communication interface 855. At least a part of the functions can also be implemented by the controller 851. For example, the controller 851 can execute the functions of the units in the electronic device 2000, so as to reduce the communication overhead of FL model training without affecting the global model accuracy.

[0189] [Application Example Regarding User Equipment]

[0190] (First Application Example)

[0191] Figure 15 FIG. is a block diagram showing an example of a schematic configuration of a smart phone 900 to which the technology of the present disclosure can be applied. The smart phone 900 includes a processor 901, a memory 902, a storage device 903, an external connection interface 904, a camera device 906, a sensor 907, a microphone 908, an input device 909, a display device 910, a speaker 911, a wireless communication interface 912, one or more antenna switches 915, one or more antennas 916, a bus 917, a battery 918, and an auxiliary controller 919.

[0192] The processor 901 can be, for example, a CPU or a system on chip (SoC), and controls the functions of the application layer and other layers of the smart phone 900. The memory 902 includes RAM and ROM, and stores data and programs executed by the processor 901. The storage device 903 can include storage media such as semiconductor memories and hard disks. The external connection interface 904 is an interface for connecting external devices (such as memory cards and universal serial bus (USB) devices) to the smart phone 900.

[0193] The camera device 906 includes image sensors (such as charge-coupled devices (CCDs) and complementary metal-oxide-semiconductor (CMOSs)), and generates captured images. The sensor 907 can include a set of sensors such as measurement sensors, gyro sensors, geomagnetic sensors, and acceleration sensors. The microphone 908 converts the sound input to the smart phone 900 into an audio signal. The input device 909 includes, for example, a touch sensor configured to detect touches on the screen of the display device 910, a keypad, a keyboard, buttons, or switches, and receives operations or information input from the user. The display device 910 includes a screen (such as a liquid crystal display (LCD) and an organic light-emitting diode (OLED) display), and displays output images of the smart phone 900. The speaker 911 converts the audio signal output from the smart phone 900 into sound.

[0194] The wireless communication interface 912 supports any cellular communication scheme (such as LTE and LTE-Advanced) and performs wireless communication. The wireless communication interface 912 generally may include, for example, a BB processor 913 and an RF circuit 914. The BB processor 913 may perform, for example, encoding / decoding, modulation / demodulation, and multiplexing / demultiplexing, and perform various types of signal processing for wireless communication. At the same time, the RF circuit 914 may include, for example, mixers, filters, and amplifiers, and transmit and receive wireless signals via an antenna 916. Note that although the figure shows a case where one RF link is connected to one antenna, this is only illustrative, and it also includes a case where one RF link is connected to multiple antennas through multiple phase shifters. The wireless communication interface 912 may be a chip module on which the BB processor 913 and the RF circuit 914 are integrated. As Figure 15 shown, the wireless communication interface 912 may include multiple BB processors 913 and multiple RF circuits 914. Although Figure 15 an example where the wireless communication interface 912 includes multiple BB processors 913 and multiple RF circuits 914 is shown, the wireless communication interface 912 may also include a single BB processor 913 or a single RF circuit 914.

[0195] In addition, in addition to the cellular communication scheme, the wireless communication interface 912 may support other types of wireless communication schemes, such as short-range wireless communication schemes, near-field communication schemes, and wireless local area network (LAN) schemes. In this case, the wireless communication interface 912 may include a BB processor 913 and an RF circuit 914 for each wireless communication scheme.

[0196] Each of the antenna switches 915 switches the connection destination of the antenna 916 among multiple circuits (such as circuits for different wireless communication schemes) included in the wireless communication interface 912.

[0197] Each of the antennas 916 includes a single or multiple antenna elements (such as multiple antenna elements included in a MIMO antenna), and is used for the wireless communication interface 912 to transmit and receive wireless signals. As Figure 15 shown, the smart phone 900 may include multiple antennas 916. Although Figure 15 an example where the smart phone 900 includes multiple antennas 916 is shown, the smart phone 900 may also include a single antenna 916.

[0198] In addition, the smart phone 900 may include an antenna 916 for each wireless communication scheme. In this case, the antenna switch 915 may be omitted from the configuration of the smart phone 900.

[0199] The bus 917 connects the processor 901, the memory 902, the storage device 903, the external connection interface 904, the imaging device 906, the sensor 907, the microphone 908, the input device 909, the display device 910, the speaker 911, the wireless communication interface 912, and the auxiliary controller 919 to each other. The battery 918 supplies power to each block of the smart phone 900 shown via a feeder line, which is partially shown as a dashed line in the figure. The auxiliary controller 919 operates the minimum necessary functions of the smart phone 900, for example, in the sleep mode. Figure 15 The battery 918 supplies power to each block of the smart phone 900 shown via a feeder line, which is partially shown as a dashed line in the figure. The auxiliary controller 919 operates the minimum necessary functions of the smart phone 900, for example, in the sleep mode.

[0200] When the electronic device 9000 is implemented as a smart phone on the user equipment side, for example Figure 15 in the case of the smart phone 900 shown, the transceiver of the electronic device 9000 can be implemented by the wireless communication interface 912. At least a part of the functions can also be implemented by the processor 901 or the auxiliary controller 919. For example, by executing the functions of the units in the above-described electronic device 9000, the processor 901 or the auxiliary controller 919 can reduce the communication overhead of FL model training without affecting the global model accuracy.

[0201] (Second application example)

[0202] Figure 16 FIG. is a block diagram showing an example of a schematic configuration of an in-vehicle navigation device 920 to which the technology of the present disclosure can be applied. The in-vehicle navigation device 920 includes a processor 921, a memory 922, a Global Positioning System (GPS) module 924, a sensor 925, a data interface 926, a content player 927, a storage medium interface 928, an input device 929, a display device 930, a speaker 931, a wireless communication interface 933, one or more antenna switches 936, one or more antennas 937, and a battery 938.

[0203] The processor 921 can be, for example, a CPU or an SoC, and controls the navigation function and other functions of the in-vehicle navigation device 920. The memory 922 includes a RAM and a ROM, and stores data and programs executed by the processor 921.

[0204] The GPS module 924 measures the position (such as latitude, longitude, and altitude) of the in-vehicle navigation device 920 using GPS signals received from GPS satellites. The sensor 925 can include a set of sensors, such as a gyro sensor, a geomagnetic sensor, and an air pressure sensor. The data interface 926 is connected to, for example, an in-vehicle network 941 via a terminal (not shown), and acquires data generated by the vehicle (such as vehicle speed data).

[0205] The content player 927 reproduces content stored in a storage medium (such as a CD and a DVD) inserted into the storage medium interface 928. The input device 929 includes, for example, a touch sensor, buttons, or switches configured to detect touches on the screen of the display device 930, and receives operations or information input from the user. The display device 930 includes a screen such as an LCD or an OLED display, and displays images of the navigation function or the reproduced content. The speaker 931 outputs sounds of the navigation function or the reproduced content.

[0206] The wireless communication interface 933 supports any cellular communication scheme (such as LTE and LTE-Advanced) and performs wireless communication. The wireless communication interface 933 generally may include, for example, a BB processor 934 and an RF circuit 935. The BB processor 934 may perform, for example, encoding / decoding, modulation / demodulation, and multiplexing / demultiplexing, and perform various types of signal processing for wireless communication. Meanwhile, the RF circuit 935 may include, for example, mixers, filters, and amplifiers, and transmit and receive wireless signals via the antenna 937. The wireless communication interface 933 may also be a single chip module on which the BB processor 934 and the RF circuit 935 are integrated. As Figure 16 shown, the wireless communication interface 933 may include multiple BB processors 934 and multiple RF circuits 935. Although Figure 16 an example in which the wireless communication interface 933 includes multiple BB processors 934 and multiple RF circuits 935 is shown, the wireless communication interface 933 may also include a single BB processor 934 or a single RF circuit 935.

[0207] In addition, in addition to the cellular communication scheme, the wireless communication interface 933 may support other types of wireless communication schemes, such as short-range wireless communication schemes, near-field communication schemes, and wireless LAN schemes. In this case, for each wireless communication scheme, the wireless communication interface 933 may include a BB processor 934 and an RF circuit 935.

[0208] Each of the antenna switches 936 switches the connection destination of the antenna 937 among multiple circuits (such as circuits for different wireless communication schemes) included in the wireless communication interface 933.

[0209] Each of the antennas 937 includes a single or multiple antenna elements (such as multiple antenna elements included in a MIMO antenna), and is used for the wireless communication interface 933 to transmit and receive wireless signals. As Figure 16 shown, the car navigation device 920 may include multiple antennas 937. Although Figure 16 an example in which the car navigation device 920 includes multiple antennas 937 is shown, the car navigation device 920 may also include a single antenna 937.

[0210] In addition, the vehicle navigation device 920 may include an antenna 937 for each wireless communication scheme. In this case, the antenna switch 936 may be omitted from the configuration of the vehicle navigation device 920.

[0211] The battery 938 supplies power to Figure 16 the respective blocks of the vehicle navigation device 920 shown via a feeder line, which is partially shown as a dashed line in the figure. The battery 938 accumulates the power supplied from the vehicle.

[0212] When the electronic device 9000 is implemented as, for example, a vehicle navigation device on the user equipment side such as Figure 16 the vehicle navigation device 920 shown, the transceiver of the electronic device 9000 may be implemented by the wireless communication interface 933. At least a part of the functions may also be implemented by the processor 921. For example, by executing the functions of the units in the above-described electronic device 9000, the processor 921 enables reduction of the communication overhead for FL model training and does not affect the global model accuracy.

[0213] The technology of the present disclosure may also be implemented as an in-vehicle system (or vehicle) 940 including one or more blocks of the vehicle navigation device 920, the in-vehicle network 941, and the vehicle module 942. The vehicle module 942 generates vehicle data (such as vehicle speed, engine speed, and fault information) and outputs the generated data to the in-vehicle network 941.

[0214] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that for those skilled in the art, all or any steps or components of the method and apparatus of the present invention can be implemented in any computing device (including a processor, a storage medium, etc.) or a network of computing devices in the form of hardware, firmware, software, or a combination thereof, which can be achieved by those skilled in the art using their basic circuit design knowledge or basic programming skills after reading the description of the present invention.

[0215] Moreover, the present invention also proposes a program product storing machine-readable instruction codes. When the instruction codes are read and executed by a machine, the method according to the embodiments of the present invention can be executed.

[0216] Correspondingly, a storage medium for carrying the above-described program product storing machine-readable instruction codes is also included in the disclosure of the present invention. The storage medium includes, but is not limited to, a floppy disk, an optical disk, a magneto-optical disk, a memory card, a memory stick, and the like.

[0217] In the case where the present invention is implemented by software or firmware, from a storage medium or a network to a computer having a dedicated hardware structure (such as Figure 17The general-purpose computer 1700 shown installs the programs that make up the software, and when various programs are installed on this computer, it can perform various functions and the like.

[0218] In Figure 17 it, the central processing unit (CPU) 1701 executes various processes according to the programs stored in the read-only memory (ROM) 1702 or the programs loaded from the storage section 1708 into the random access memory (RAM) 1703. In the RAM 1703, data required when the CPU 1701 executes various processes and the like is also stored as needed. The CPU 1701, ROM 1702, and RAM 1703 are connected to each other via a bus 1704. The input / output interface 1705 is also connected to the bus 1704.

[0219] The following components are connected to the input / output interface 1705: an input section 1706 (including a keyboard, a mouse, etc.), an output section 1707 (including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.), a storage section 1708 (including a hard disk, etc.), a communication section 1709 (including a network interface card such as a LAN card, a modem, etc.). The communication section 1709 performs communication processing via a network such as the Internet. As needed, a drive 1710 may also be connected to the input / output interface 1705. A removable medium 1711 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is installed on the drive 1710 as needed, so that the computer programs read therefrom are installed into the storage section 1708 as needed.

[0220] In the case where the above-described series of processes are implemented by software, the programs that make up the software are installed from a network such as the Internet or a storage medium such as the removable medium 1711.

[0221] Those skilled in the art should understand that such a storage medium is not limited to Figure 17 the removable medium 1711 shown in which programs are stored and distributed separately from the device to provide the programs to users. Examples of the removable medium 1711 include magnetic disks (including floppy disks (registered trademark)), optical disks (including compact disc read-only memories (CD-ROMs) and digital versatile discs (DVDs)), magneto-optical disks (including mini discs (MD) (registered trademark)), and semiconductor memories. Alternatively, the storage medium may be the ROM 1702, the hard disk included in the storage section 1708, etc., in which programs are stored and distributed to users together with the devices containing them.

[0222] It should also be noted that in the devices, methods, and systems of the present invention, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent solutions of the present invention. Moreover, the steps of performing the above series of processes can naturally be executed in chronological order according to the described order, but it is not necessary to be executed in chronological order. Certain steps can be executed in parallel or independently of each other.

[0223] Finally, it should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. In addition, without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0224] Although the embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings, it should be understood that the above-described embodiments are only used to illustrate the present invention and do not constitute a limitation to the present invention. For those skilled in the art, various modifications and changes can be made to the above embodiments without departing from the essence and scope of the present invention. Therefore, the scope of the present invention is only defined by the appended claims and their equivalent meanings.

[0225] The present technology can also be implemented as follows.

[0226] Solution 1. An electronic device for a federated learning server, comprising:

[0227] At least one processor; and

[0228] At least one memory, including computer program code, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to execute:

[0229] Send magnitude information about the parameter magnitude of the global model for federated learning to the core network to assist the core network in identifying, from among multiple federated learning clients, a federated learning client that cannot upload all the parameters of its local model as a partial upload client, and

[0230] Based on the partial parameters of the local model received from the partial upload client, perform aggregation under the condition of meeting a predetermined condition to obtain the global model in the current round of training of federated learning.

[0231] Solution 2. The electronic device according to Solution 1, wherein the magnitude information includes information about the size of the parameters of the global model.

[0232] Solution 3. The electronic device according to Solution 1 or 2, wherein,

[0233] the predetermined condition includes receiving, from the core network, a list of the partial upload clients and / or the accuracy of the global model in the previous round of training reaching a preset accuracy, wherein the list includes the IDs of the partial upload clients, and

[0234] in the case where the accuracy of the global model in the previous round of training reaches the preset accuracy, all the multiple federated learning clients serve as the partial upload clients.

[0235] Solution 4. The electronic device according to Solution 3, wherein,

[0236] the IDs of the partial upload clients included in the list are determined by the core network based on the magnitude information and the channel information predicted between the electronic device and the federated learning clients.

[0237] Solution 5. The electronic device according to Solution 4, wherein,

[0238] the channel information includes at least one of a channel quality indicator CQI, a precoding matrix indicator PMI, a reference signal received power RSRP, and a signal-to-noise ratio SNR of the channel between the electronic device and the federated learning clients.

[0239] Solution 6. The electronic device according to any one of Solutions 1 to 5, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform:

[0240] For the parameters of the global model in the previous round of training and the parameters of the local models uploaded by the multiple federated learning clients in the previous round of training:

[0241] calculate a local model parameter deviation degree representing the degree of dispersion between the parameters of the local models of the multiple federated learning clients, and calculate a global model parameter deviation degree representing the degree of deviation between the parameters of the local models of the multiple federated learning clients and the parameters of the global model, and

[0242] determine the partial parameters based on the local model parameter deviation degree and the global model parameter deviation degree.

[0243] Solution 7. The electronic device according to Solution 6, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform:

[0244] Calculate the average value of the i-th parameter based on the i-th parameters of the local models of the multiple federated learning clients, and calculate the deviation degree of the local model parameters for the i-th parameter based on the difference between the i-th parameter of each local model of the federated learning client and the average value; and

[0245] Calculate the deviation degree of the global model parameters for the i-th parameter based on the difference between the i-th parameter of each local model of the federated learning client and the i-th parameter of the global model.

[0246] Solution 8. The electronic device according to Solution 6 or 7, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform:

[0247] Based on the local model parameter deviation degree and the global model parameter deviation degree, use a clustering method to cluster the parameters, thereby obtaining multiple clusters after clustering,

[0248] Select the selected clusters that meet the predetermined cluster conditions from the multiple clusters, and

[0249] Use the parameters in the selected clusters as the partial parameters.

[0250] Solution 9. The electronic device according to Solution 8, wherein,

[0251] The predetermined cluster condition includes that the selected probability corresponding to the cluster is greater than a predetermined probability, and

[0252] The at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform:

[0253] Respectively take the local model parameter deviation degree and the global model parameter deviation degree as the coordinates of a two-dimensional coordinate system, and use the local model parameter deviation degree and the global model parameter deviation degree corresponding to the calculated parameters as coordinate values, so as to represent each parameter as a point in the coordinate system,

[0254] Use the K-Means algorithm to cluster all the points in the coordinate system, thereby obtaining the multiple clusters, and

[0255] For each of at least some of the multiple clusters, calculate the selected probability of the cluster based on the distance between the cluster center of the cluster and the origin of the coordinates, and the angle between the line connecting the cluster center to the origin of the coordinates and the dividing line of the first quadrant of the coordinate system, wherein the dividing line is a straight line passing through the origin of the coordinates and making a predetermined angle with the horizontal axis of the first quadrant,

[0256] Wherein, the greater the distance, the greater the probability of being selected, and the smaller the included angle, the greater the probability of being selected.

[0257] Solution 10. The electronic device according to any one of Solutions 1 to 9, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform:

[0258] Send notification information about the partial parameters to the partial upload client, for the partial upload client to only upload the partial parameters in the current round of training.

[0259] Solution 11. The electronic device according to Solution 10, wherein

[0260] The global model is a hierarchical neural network model,

[0261] The notification information includes the position information of the partial parameters in the neural network model.

[0262] Solution 12. The electronic device according to Solution 11, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform:

[0263] Based on the position information, align the partial parameters of the local model received from the partial upload client and the all parameters of the local models received from other clients among the multiple federated learning clients, and perform aggregation of the local models.

[0264] Solution 13. The electronic device according to any one of Solutions 1 to 12, wherein the magnitude information is used as an input for UE communication analysis in the network data analysis function NWDAF, and the UE communication prediction output of the UE communication analysis includes a list with the IDs of the partial upload clients.

[0265] Solution 14. An electronic device for a core network, comprising:

[0266] At least one processor; and

[0267] At least one memory, including computer program code, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform:

[0268] Receive magnitude information regarding the order of magnitude of the parameters of the global model for federated learning from a federated learning server to identify, from among a plurality of federated learning clients, a federated learning client that cannot upload all the parameters of its local model as a partial upload client, so that, when a predetermined condition is met, the federated learning server aggregates based on partial parameters of the local model received from the partial upload client to obtain the global model in the current round of training of federated learning.

[0269] Solution 15. The electronic device according to Solution 14, wherein the magnitude information includes information regarding the size of the parameters of the global model.

[0270] Solution 16. The electronic device according to Solution 14 or 15, wherein

[0271] the predetermined condition includes that the federated learning server receives a list of the partial upload clients and / or the accuracy of the global model in the previous round of training reaches a preset accuracy from the electronic device, wherein the list includes the IDs of the partial upload clients.

[0272] Solution 17. The electronic device according to Solution 16, wherein the at least one memory and the computer program code are configured to, via the at least one processor, cause the electronic device to perform:

[0273] Determine the IDs of the partial upload clients included in the list based on the magnitude information and the predicted channel information between the federated learning server and the federated learning client.

[0274] Solution 18. The electronic device according to Solution 17, wherein

[0275] the channel information includes at least one of a channel quality indicator (CQI), a precoding matrix indicator (PMI), a reference signal received power (RSRP), and a signal-to-noise ratio (SNR) of the channel between the federated learning server and the federated learning client.

[0276] Solution 19. The electronic device according to any one of Solutions 14 to 18, wherein the magnitude information is used as an input for UE communication analysis in a network data analytics function (NWDAF), and the UE communication prediction output of the UE communication analysis includes a list having the IDs of the partial upload clients.

[0277] Solution 20. An electronic device for a federated learning client, comprising:

[0278] At least one processor; and

[0279] At least one memory, including computer program code, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform:

[0280] When a predetermined condition is met, send partial parameters of a local model for federated learning to a federated learning server for the federated learning server to aggregate the local models based on the partial parameters to obtain a global model in the current round of training of the federated learning,

[0281] wherein the federated learning server sends magnitude information about the parameter magnitude of the global model to the core network to assist the core network in identifying the electronic device as a partial upload client that cannot upload all the parameters of its local model.

[0282] Solution 21. The electronic device according to Solution 20, wherein the magnitude information includes information about the size of the parameters of the global model.

[0283] Solution 22. The electronic device according to Solution 20 or 21, wherein,

[0284] the predetermined condition includes that the federated learning server receives from the core network a list including the ID of the partial upload client and / or the accuracy of the global model in the previous round of training reaches a preset accuracy.

[0285] Solution 23. The electronic device according to any one of Solutions 20 to 22, wherein the at least one memory and the computer program code are configured to, through the at least one processor, cause the electronic device to perform:

[0286] Receive notification information about the partial parameters from the federated learning server.

[0287] Solution 24. The electronic device according to Solution 23, wherein,

[0288] the global model is a hierarchical neural network model,

[0289] the notification information includes position information of the partial parameters in the neural network model.

[0290] Solution 25. The electronic device according to any one of Solutions 20 to 24, wherein the magnitude information is used as an input for UE communication analysis in a network data analytics function NWDAF, and the UE communication prediction output of the UE communication analysis includes a list with the ID of the partial upload client.

[0291] Solution 26. A method for a federated learning server, including:

[0292] Send magnitude information about the parameter magnitude of the global model for federated learning to the core network to assist the core network in identifying, among multiple federated learning clients, a federated learning client that cannot upload all the parameters of its local model as a partial upload client, and

[0293] When a predetermined condition is satisfied, aggregate based on the partial parameters of the local model received from the partial upload client to obtain the global model in the current round of training of federated learning.

[0294] Solution 27. A method for a core network, including:

[0295] Receive magnitude information about the parameter magnitude of the global model for federated learning from a federated learning server to identify, among multiple federated learning clients, a federated learning client that cannot upload all the parameters of its local model as a partial upload client, so that when a predetermined condition is satisfied, the federated learning server aggregates based on the partial parameters of the local model received from the partial upload client to obtain the global model in the current round of training of federated learning.

[0296] Solution 28. A method for a federated learning client, including:

[0297] When a predetermined condition is satisfied, send the partial parameters of the local model for federated learning to the federated learning server for the federated learning server to aggregate the local models based on the partial parameters to obtain the global model in the current round of training of federated learning,

[0298] wherein the federated learning server sends magnitude information about the parameter magnitude of the global model to the core network to assist the core network in identifying the federated learning client as a partial upload client that cannot upload all the parameters of its local model.

[0299] Solution 29. A computer-readable storage medium, on which computer-executable instructions are stored, and when the computer-executable instructions are executed, the method according to any one of Solutions 26 to 28 is executed.

Claims

1. An electronic device for a federated learning server, comprising: at least one processor; and at least one memory including computer program code, wherein the at least one memory and the computer program code are configured to, by means of the at least one processor, cause the electronic device to perform: send to a core network magnitude information regarding the order of magnitude of the number of parameters of a global model for federated learning, to assist the core network in identifying, from among a plurality of federated learning clients, a federated learning client that cannot upload all parameters of its local model as a partial upload client, and when a predetermined condition is satisfied, aggregate based on partial parameters of a local model received from the partial upload client to obtain a global model in the current round of training of federated learning.

2. The electronic device according to claim 1, wherein, The magnitude information includes information regarding the size of the parameters of the global model.

3. The electronic device according to claim 1 or 2, wherein, The predetermined condition includes receiving from the core network a list regarding the partial upload clients and / or the accuracy of the global model in the previous round of training reaching a preset accuracy, wherein the list includes the IDs of the partial upload clients, and when the accuracy of the global model in the previous round of training reaches the preset accuracy, all of the plurality of federated learning clients serve as the partial upload clients.

4. The electronic device according to claim 3, wherein, The IDs of the partial upload clients included in the list are determined by the core network based on the magnitude information and predicted channel information between the electronic device and the federated learning clients.

5. An electronic device for a core network, comprising: at least one processor; and at least one memory including computer program code, wherein the at least one memory and the computer program code are configured to, by means of the at least one processor, cause the electronic device to perform: receive from a federated learning server magnitude information regarding the order of magnitude of the number of parameters of a global model for federated learning, to identify, from among a plurality of federated learning clients, a federated learning client that cannot upload all parameters of its local model as a partial upload client, for the federated learning server to aggregate based on partial parameters of the local model received from the partial upload client to obtain a global model in the current round of training of federated learning when a predetermined condition is satisfied.

6. An electronic device for a federated learning client, comprising: at least one processor; and at least one memory including computer program code, wherein the at least one memory and the computer program code are configured to, by means of the at least one processor, cause the electronic device to perform: when a predetermined condition is satisfied, send to a federated learning server partial parameters of a local model for federated learning, for the federated learning server to aggregate the local model based on the partial parameters to obtain a global model in the current round of training of federated learning, wherein the federated learning server sends to a core network magnitude information regarding the order of magnitude of the number of parameters of a global model, to assist the core network in identifying the electronic device as a partial upload client that cannot upload all parameters of its local model.

7. A method for a federated learning server, comprising: send to a core network magnitude information regarding the order of magnitude of the number of parameters of a global model for federated learning, to assist the core network in identifying, from among a plurality of federated learning clients, a federated learning client that cannot upload all parameters of its local model as a partial upload client, and Under the condition of meeting the predetermined conditions, aggregate based on the partial parameters of the local model received from the partial upload client to obtain the global model in the current round of training of federated learning.

8. A method for a core network, comprising: Receive magnitude information about the parameter magnitude of the global model for federated learning from the federated learning server to identify, among multiple federated learning clients, the federated learning clients that cannot upload all the parameters of their local models as partial upload clients, so that, under the condition of meeting the predetermined conditions, the federated learning server aggregates based on the partial parameters of the local model received from the partial upload clients to obtain the global model in the current round of training of federated learning.

9. A method for a federated learning client, comprising: Under the condition of meeting the predetermined conditions, send the partial parameters of the local model for federated learning to the federated learning server for the federated learning server to aggregate the local models based on the partial parameters to obtain the global model in the current round of training of federated learning. Wherein, the federated learning server sends magnitude information about the parameter magnitude of the global model to the core network to assist the core network in identifying the federated learning client as a partial upload client that cannot upload all the parameters of its local model.

10. A computer-readable storage medium having computer-executable instructions stored thereon, which when executed, perform the method according to any one of claims 7 to 9.

Images