Electronic equipment for wireless communication

Abstract

This disclosure relates to electronic equipment and methods for wireless communication, and computer-readable storage media. The electronic equipment for wireless communication includes a processing circuit configured to determine, in conjunction with user equipment within the service range of the electronic equipment, whether a user equipment participates in federated learning non-independently by dividing a model to be trained and training the resulting divided models in cooperation with other user equipment, or independently by training the model to be trained independently.

Description

[Technical Field]

[0001] This application claims priority to a Chinese patent application filed with the China National Intellectual Property Office on March 31, 2023, with application number 202310342408.6, and titled "Electronic device and method for wireless communication, computer-readable storage medium," and incorporates its entire contents into this application by reference.

[0002] This disclosure relates to the field of wireless communication technology, specifically to electronic equipment and methods for wireless communication, and computer-readable storage media. More specifically, it relates to more efficient methods for performing federated learning. [Background technology]

[0003] When performing Federated Learning (FL), the user's device uploads its locally trained model to a base station or server to perform model aggregation. In some cases (for example, if the user's device lacks sufficient computing or communication capabilities), federated learning may not be able to be performed efficiently.

[0004] How to implement associative learning more efficiently is a current focus of research. [Overview of the project] [Means for solving the problem]

[0005] The following provides a brief overview of the present invention to offer a basic understanding of certain aspects of it. It should be understood that this overview is not exhaustive. It is not intended to identify any essential or important parts of the invention, nor to intentionally limit its scope. Its purpose is to provide a simplified concept in order to precede the more detailed description that will follow.

[0006] According to one aspect of this disclosure, the present invention provides an electronic device for wireless communication, which includes a processing circuit configured to determine, in federated learning, whether a user device will participate in federated learning non-independently by training a divided model obtained by dividing the model to be trained in cooperation with other user devices, or whether it will participate in federated learning independently by training the model to be trained independently, based on self-state information reported by a user device within the service range of the electronic device.

[0007] In the embodiments of this disclosure, the electronic device can perform associative learning more efficiently by deciding whether the user device participates in associative learning independently or independently.

[0008] According to one aspect of this disclosure, the present invention provides an electronic device for wireless communication, which includes a processing circuit configured such that a network-side device providing services to the electronic device reports the electronic device's self-state information to the network-side device so that the network-side device can determine whether, in federated learning, the electronic device will participate in federated learning non-independently by training the divided models obtained by dividing the model to be trained in cooperation with other electronic devices, or whether it will participate in federated learning independently by training the model to be trained independently.

[0009] In the embodiments of this disclosure, the electronic device can perform federated learning more efficiently by reporting its own state information to the network-side device, so that the network-side device can determine whether the electronic device participates in federated learning independently or independently.

[0010] According to one aspect of this disclosure, the present invention provides an electronic device for wireless communication, which includes a processing circuit configured to report the self-state information of the electronic device to a network-side device that provides services to the electronic device, so that the network-side device can determine, in federated learning, whether the electronic device can divide the model to be trained and train the resulting divided models in cooperation with other electronic devices, and whether the other electronic devices will participate in federated learning independently or whether the other electronic devices will train the model to be trained independently and participate in federated learning independently.

[0011] In the embodiments of this disclosure, the network-side device can determine whether other electronic devices participate in federated learning independently or independently, and electronic devices can perform federated learning more efficiently by reporting their own state information to the network-side device.

[0012] According to one aspect of this disclosure, a wireless communication system including the above-mentioned electronic equipment is provided.

[0013] In one aspect of this disclosure, a method for wireless communication is provided, which includes determining, in federated learning, whether a user device participates in federated learning non-independently by training a divided model obtained by dividing a model to be trained in cooperation with other user devices, or independently by training a model to be trained independently, based on self-state information reported by a user device within the service range of an electronic device.

[0014] According to one aspect of the present disclosure, there is provided a method for wireless communication. In the method, a network-side device that provides services to an electronic device determines, in federated learning, whether the electronic device participates in federated learning non-independently by training a split model obtained by splitting a training target model in cooperation with other electronic devices, or participates in federated learning independently by training the training target model independently. The method includes reporting the self-state information of the electronic device to the network-side device.

[0015] According to one aspect of the present disclosure, there is provided a method for wireless communication. In the method, a network-side device that provides services to an electronic device determines, in federated learning, whether another electronic device participates in federated learning non-independently by training a split model obtained by splitting a training target model in cooperation with other electronic devices, or participates in federated learning independently by training the training target model independently. The method includes reporting the self-state information of the electronic device to the network-side device.

[0016] According to another aspect of the present invention, there are further provided computer program code for realizing the above method for wireless communication, a computer program product, and a computer-readable storage medium on which the computer program code for realizing the above method for wireless communication is recorded.

Brief Description of the Drawings

[0017] To further explain the above and other advantages and features of the present invention, the specific embodiments of the present invention will be described in more detail below in conjunction with the drawings. The drawings are included in this specification together with the following detailed description and form a part of this specification. Elements having the same function and configuration are denoted by the same reference numerals. It should be noted that these drawings illustrate only typical examples of the present invention and should not be regarded as a limitation on the scope of the present invention. In the drawings,

[0018] [Figure 1]FIG. 1 shows a functional block diagram of an electronic device for wireless communication according to an embodiment of the present disclosure. [Figure 2] FIG. 2 is a diagram showing an example of a federated learning network. [Figure 3] FIG. 3 is a diagram showing an example of split learning according to an embodiment of the present disclosure. [Figure 4] FIG. 4 is a diagram showing examples of different split points according to an embodiment of the present disclosure. [Figure 5] FIG. 5 is a diagram showing a configuration example of a federated learning network according to an embodiment of the present disclosure. [Figure 6] FIG. 6 shows a functional block diagram of an electronic device for wireless communication according to another embodiment of the present disclosure. [Figure 7] FIG. 7 shows a functional block diagram of an electronic device for wireless communication according to yet another embodiment of the present disclosure. [Figure 8] FIG. 8 shows a flowchart of a method for wireless communication according to an embodiment of the present disclosure. [Figure 9] FIG. 9 shows a flowchart of a method for wireless communication according to another embodiment of the present disclosure. [Figure 10] FIG. 10 shows a flowchart of a method for wireless communication according to yet another embodiment of the present disclosure. [Figure 11] FIG. 11 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. [Figure 12] FIG. 12 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. [Figure 13] FIG. 13 is a block diagram showing an example of a schematic configuration of a smartphone to which the technology of the present disclosure can be applied. [Figure 14] FIG. 14 is a block diagram showing an example of a schematic configuration of a car navigation device to which the technology of the present disclosure can be applied. [Figure 15] FIG. 15 is a block diagram showing a schematic configuration of a general-purpose personal computer that can implement a method and / or apparatus and / or system according to an embodiment of the present disclosure. [Modes for carrying out the invention]

[0019] The following describes exemplary embodiments of the present invention, accompanied by drawings. For clarity and brevity, not all features of actual embodiments are described in this specification. It should be understood that decisions to specify such actual embodiments must be made during the development process to achieve the specific goals of the developers, such as those relating to system and business limitations, which may vary depending on the embodiment. It should also be understood that while development work can be very complex and time-consuming, such development work is routine for those skilled in the art who would benefit from this disclosure.

[0020] It should also be noted that, in order to avoid obscuring the present invention with unnecessary details, the drawings show only the apparatus configuration and / or processing steps closely related to the solution of the present invention, and omit other details that are not significantly related to the present invention.

[0021] Figure 1 shows a functional block diagram of an electronic device 100 for wireless communication according to one embodiment of the present disclosure.

[0022] As shown in Figure 1, the electronic device 100 includes a processing unit 101 that, in conjunction with user devices within the service scope of the electronic device 100, can determine whether a user device will participate in federated learning non-independently by dividing (segmenting) the model to be trained and training the resulting segmented models in cooperation with other user devices, or whether it will participate in federated learning independently by independently training the model to be trained.

[0023] The processing unit 101 may be implemented by one or more processing circuits, and these processing circuits may be implemented, for example, as a chip.

[0024] The electronic device 100, as a network-side device in a wireless communication system, can be specifically installed, for example, on the base station side, or communicated with a base station. Here, the electronic device 100 may be implemented at the chip level or at the device level. For example, the electronic device 100 may operate as the base station itself, and may further include external devices such as memory and transceivers (not shown). The memory is used to store programs executed by the base station to perform various functions, and related data information. The transceiver may include one or more communication interfaces to support communication between different devices (e.g., user equipment (UE), other base stations, etc.), but the implementation of the transceiver is not specifically limited here.

[0025] For example, the network-side equipment may be a base station, and the base station may be, for example, an eNB or a gNB.

[0026] For example, the network-side device may be a server.

[0027] The wireless communication system described herein may be a 5G NR (New Radio) communication system. Furthermore, the wireless communication system described herein may include a non-terrestrial network (NTN). Optionally, the wireless communication system described herein may also include a terrestrial network (TN). Those skilled in the art will also understand that the wireless communication system described herein may be a 4G or 3G communication system.

[0028] For example, the model to be trained may be a deep learning model, and may be at least one of the following: RNN (Recurrent Neural Network), CNN (Convolutional Neural Network), and ANN (Artificial Neural Network). In the following text, for simplicity, the model to be trained may be abbreviated as "model."

[0029] For example, the user device estimates its own state information Info U and reports it to the electronic device 100.

[0030] FIG. 2 is a diagram showing an example of a federated learning network.

[0031] As shown in FIG. 2, it is assumed that there are K (where K is a positive integer greater than 1) user devices (Device 1, Device 2, Device 3,..., Device K-1, Device K). The federated learning process of the federated learning network is as follows. (1) Each user device accesses a server (Server) or a base station (hereinafter, for simplicity, the base station is described as an example) via a wireless channel, and obtains the parameters (W 0 ) of the initial global model as the parameters of the initial local model, respectively, and W1 0 =W2 0 =…=W k 0 =W 0 , where W K 0 represents the parameters of the initial local model of device k (k = 1, 2,..., K). Hereinafter, for simplicity, the parameters of the global model may be abbreviated as the global model, and the parameters of the local model may be abbreviated as the local model. (2) Each user device performs learning using the data stored locally, completes one iterative update (local update) of the local model,

Number

number

number

[0032] For example, the training model described above corresponds to either the initial global model or the updated global model in Figure 2.

[0033] User devices that participate in federative learning independently can be called standalone members in federative learning. Non-standalone members are also defined in federative learning, and include UEs that participate in federative learning independently, as well as other user devices that interact with such UEs. Other user devices that constitute a UE and a Non-Standalone Member may be called Assistant Equipment (AE). For example, an Assistant Equipment may be a peripheral device of a UE. Hereinafter, Standalone Members and Non-Standalone Members may be collectively referred to as Members.

[0034] Figure 3 shows an example of segmented learning according to an embodiment of this disclosure. In segmented learning, the model to be trained is divided to obtain segmented models, and these segmented models are then trained.

[0035] First, we establish the model topology. As shown in Figure 3, the complete model to be trained is divided into two parts, for example, at the split point: the earlier model located on the left (the first split model) and the later model located on the right (the second split model). We assume that the UE trains the first split model and the AE trains the second split model. For simplicity, we can abbreviate the first split model as the UE model and the second split model as the AE model. That is, the complete model to be trained is split between the UE and the AE, becoming the UE model and the AE model. Note that although Figure 3 shows the model being divided into two parts, those skilled in the art will understand that the model can be divided into any other integer number of parts.

[0036] At the start of training, the training parameters for the UE model and AE model are initialized randomly.

[0037] During the training process, the UE performs forward computation on the UE model based on local data and sends the local label (Label) and intermediate data #1 obtained from the forward computation (e.g., the output tensor of the UE model) to the AE. After receiving intermediate data #1, the AE continues forward computation on the AE model and performs backward gradient computation according to the uploaded label to obtain the gradient corresponding to the AE model and updates the parameters of the AE model according to the obtained gradient. The AE sends intermediate data #2, for example, the gradient of the AE model, to the UE in the backward direction, and the UE continues backward computation to obtain the gradient corresponding to the UE model. The UE updates the parameters of the UE model according to the gradient corresponding to the UE model. This process is repeated until the model converges. In summary: 1) The UE first runs the UE model (forward propagation algorithm) and sends the obtained intermediate data #1 to the AE. 2) The AE uses the received intermediate data #1 as input to the AE model, executes the forward propagation algorithm to obtain intermediate results, and then, according to the intermediate results, executes the backward propagation algorithm to obtain intermediate data #2 (e.g., the gradient of the AE model) and updates the AE model. AE sends intermediate data #2 to UE in the reverse direction. 3) UE updates its UE model by executing a reverse propagation algorithm based on intermediate data #2. Steps 1) to 3) may be repeated several times in each round of global training in associative learning.

[0038] In the case of broadcasting, the UE can improve the privacy of its local data by first performing one or more full local model updates, and then sending the updated AE model to the AE for split learning.

[0039] If a user device is a Standalone Member, it does not participate in segmented learning. Instead, it independently trains the complete target model locally, and the trained local model then performs federative learning by executing global model aggregation together with the trained local models of other Members. Alternatively, this user device can configure itself as a Non-Standalone Member with the above-mentioned AE, perform segmented learning on the target model locally, and the trained local model of this Non-Standalone Member then performs federative learning by executing global model aggregation together with the trained local models of other Members.

[0040] Both Standalone Members and Non-Standalone Members can be considered equivalent to clients in conventional federative learning networks. However, compared to clients in conventional federative learning networks, Non-Standalone Members include both UEs and AEs, and by dividing the model to be trained, a portion of the model that would normally be trained on the UE side is placed on the AE side and trained by the AE.

[0041] In the embodiments of this disclosure, the electronic device 100 can perform federated learning more efficiently by determining whether user devices participate in federated learning independently or independently. For example, the computing or communication capabilities of different UEs participating in federated learning may differ; some UEs may have important data but lack computing or communication capabilities, while others may have strong computing and communication capabilities. For example, a UE with insufficient computing or communication capabilities can reduce the load on the UE participating in the federated learning process by training a segmented model in cooperation with other user devices (AEs). A UE with strong computing and communication capabilities can independently complete training on the model to be trained locally without coordinating with AEs. It can also fully utilize the computing or communication resources of AEs in the communication network. Furthermore, since the number of users participating in federated learning that the electronic device 100 can generally support is limited, it is necessary to select UEs before starting federated learning. Some devices that have not been selected but wish to participate in (contribute to) federated learning can participate in training as AEs. In the case of a Non-Standalone Member, the AE only holds parameters for some of the models of the training target model, and by sending intermediate data between the UE and AE, data privacy and security are ensured, i.e., data privacy protection is further enhanced.

[0042] For example, an AE (Environmental Exposure) may be a device with fairly strong computing capabilities in a communication network. For instance, if a mobile phone is the UE (Unified User), a vehicle or drone (UAV) may be the AE; if a vehicle is the UE, a Roadside Unit (RSU) may be the AE.

[0043] For example, the user device's self-state information may include first channel state information of the link between the user device and the electronic device 100, and the processing unit 101 may be configured to decide that the user device will participate in federated learning independently if the value of the first channel state information is smaller than a first predetermined threshold.

[0044] For example, a person skilled in the art can set a first predetermined threshold depending on their experience or application scenario.

[0045] As an example, the first channel state information includes at least one of the following: signal-to-interference noise ratio (SINR), reference signal received power (RSRP), reference signal received quality (RSRQ), and reference signal received strength (RSRI) of the uplink between the user equipment and the electronic equipment 100.

[0046] Taking SINR as an example, if the SINR of the uplink is less than a first predetermined threshold, the user device and other user devices decide to become Non-Standalone Members and participate in federated learning.

[0047] As an example, the user device's self-state information includes the user device's computing and storage capability information, and the processing unit 101 can be configured to decide that the user device will participate in federative learning non-independently if the size of the storage space indicated by the computing and storage capability information is smaller than the size of the model to be trained, and / or the computing capability indicated by the computing and storage capability information is smaller than the computing capability required to train the model to be trained. In other words, in the above case, the user device and other user devices decide to form Non-Standalone Members and participate in federative learning.

[0048] For example, computing and storage capability information includes at least one of the following: the CPU usage rate of the user's device and the memory size.

[0049] For example, the user device's self-state information includes the user device's local data information, and the processing unit 101 can be configured to decide that the user device will participate in federative learning non-independently if the training time required for the user device to train the training target model, obtained based on the local data information and the user device's computing and storage capability information, is greater than a second predetermined threshold. That is, if the training time is greater than the second predetermined threshold, the user device and other user devices decide to form Non-Standalone Members and participate in federative learning.

[0050] For example, a person skilled in the art can set a second predetermined threshold depending on their experience or application scenario.

[0051] For example, local data information includes the number of training samples used when user devices independently participate in federative learning, and the sample dimensions.

[0052] For example, the user device's self-state information may include the user device's power consumption information, and the processing unit 101 may be configured to decide that the user device will participate in federated learning independently if the power consumption indicated by the power consumption information is less than a third predetermined threshold. That is, if the power consumption indicated by the power consumption information is less than a third predetermined threshold, the user device and other user devices decide to form Non-Standalone Members and participate in federated learning.

[0053] For example, a person skilled in the art can set a third predetermined threshold depending on their experience or application scenario.

[0054] For example, the user device's self-state information may include the user device's location information, and the processing unit 101 may be configured to decide whether the user device should participate in federated learning independently if it determines, based on the location information, that no other user devices are within a predetermined range of the user device. That is, if it is determined that no other user devices are within a predetermined range of the user device, it is decided that the user device should participate in federated learning as a Standalone Member.

[0055] For example, a person skilled in the art can set a predetermined range depending on their experience or application scenario.

[0056] For example, the user device's self-state information may include the user device's movement information, and the processing unit 101 may be configured to decide whether the user device should participate in federated learning independently if it determines, based on the movement information, that the user device has high mobility. That is, if the user device has high mobility, it is decided that the user device should participate in federated learning as a Standalone Member.

[0057] For example, movement information includes at least one of the following: the user's device's speed, direction of movement, and duration of stay.

[0058] For example, the processing unit 101 can be configured to select a collaborating user device to train with the user device based on second channel state information of sidelinks between user devices within a predetermined range of the user device, when the user device decides to participate in associative learning independently. In other words, the collaborating user device is not fixed for the user device. For example, the collaborating user device is the AE described above.

[0059] For example, the processing unit 101 can be configured to select other user devices as linked user devices if the value of the second channel status information of the side link corresponding to the other user device is greater than a fourth predetermined threshold (which is AE selection condition 1).

[0060] For example, a person skilled in the art can set a fourth predetermined threshold depending on their experience or application scenario.

[0061] As an example, the second channel state information includes at least one of the following: sidelink signal-to-interference noise ratio (SINR), reference signal received power (RSRP), reference signal received quality (RSRQ), and reference signal received strength (RSRI).

[0062] As an example, the processing unit 101 can be configured to select as a linked user device other user devices that satisfy the condition (AE selection condition 2) that the user device (UE) can transmit model data relating to the partition model corresponding to the user device to the other user device within a first predetermined time range, and / or that the other user device can transmit model data relating to the partition model corresponding to the other user device to the user device within a second predetermined time range, via a side link corresponding to the other user device.

[0063] For example, the model data for the partitioned model corresponding to UE may be intermediate data #1 as explained with reference to Figure 3, and the model data for the partitioned model corresponding to AE may be intermediate data #2 as explained with reference to Figure 3.

[0064] For example, a person skilled in the art can set a first predetermined time range (e.g., indicated by t1) and / or a second predetermined time range (e.g., indicated by t2) depending on their experience or application scenario.

[0065] As an example, the processing unit 101 can be configured to determine a split point for splitting the training target model to obtain a split model, based on second channel state information, when a target user device (AE) is selected. The above split point may include one or more split points. Although the description here has described selecting the target user device (AE) and then determining the split point, it is also possible to select the target user device (AE) after determining the split point. In this case, the selection conditions may include, for example, the above-mentioned AE selection condition 1 and / or AE selection condition 2.

[0066] For example, in the case of a Non-Standalone Member, the dividing point between the UE and AE is predetermined by the electronic device 100 based on second channel state information.

[0067] Figure 4 shows examples of different division points according to embodiments of the present disclosure.

[0068] We assume that the model to be trained (e.g., a neural network) has M layers, and m indicates the location of the split points. m=0 indicates that the model to be trained is completely offloaded to the AE (the model is fully trained by the AE), and m=M indicates that the model to be trained is not split and is fully trained by the UE. Figure 4 shows an example where M=5.

[0069] In the case of a Non-Standalone Member, the following applies: 1). When m = 0, the training target model is completely offloaded to the AE. The AE does not need to have the data, and the UE needs to send the original data to the AE, and send it only once during the entire training process (for example, when there is no additional notification from the electronic device 100). After the AE completes the local training, the AE sends the trained model to the UE, and the UE can upload the trained local model to the electronic device 100 (for example, when the uplink state of the UE is good and the uplink state of the AE is bad). The AE can also directly upload the model to the electronic device 100. 2). When m = M, the training target model needs to be completely trained on the UE. In this case, the AE functions only as a relay. 3). When 0 < m < M, the UE and the AE send intermediate data to each other via sidelink to perform split learning.

[0070] As an example, the processing unit 101 can be configured such that, via a sidelink between the user equipment (UE) and the cooperating user equipment (AE), the condition that the UE can send model data regarding the split model corresponding to the UE to the cooperating user equipment within a first predetermined time range, and / or the cooperating user equipment can send model data regarding the split model corresponding to the cooperating user equipment to the UE within a second predetermined time range, is satisfied by the determined split point. For example, the model data regarding the split model corresponding to the UE can be the intermediate data #1 described with reference to FIG. 3, and the model data regarding the split model corresponding to the AE can be the intermediate data #2 described with reference to FIG. 3.

[0071] For example, the self-state information of a user device includes the second channel state information of sidelinks between user devices within a predetermined range of the user device. Since the sidelink between the UE and AE changes over time, dynamic decisions regarding associative learning must be made in response to the changes in the sidelink. For example, whether the UE participates in associative learning as a Standalone Member or a Non-standalone Member, which AE the UE forms a Non-standalone Member with, and the location of the split point must all be dynamically determined in response to the changes in the sidelink.

[0072] For example, the processing unit 101 can be configured to decide that a user device will participate in associative learning independently if there are other user devices within a predetermined range whose second channel state information value is greater than a fifth predetermined threshold.

[0073] For example, a person skilled in the art may set a fifth predetermined threshold depending on their experience or application scenario.

[0074] For example, the processing unit 101 can be configured to decide that a user device will independently participate in federated learning if the second channel state information corresponding to other user devices within a predetermined range is all below a fifth predetermined threshold.

[0075] As an example, the second channel state information includes at least one of the following: sidelink signal-to-interference noise ratio (SINR), reference signal received power (RSRP), reference signal received quality (RSRQ), and reference signal received strength (RSRI).

[0076] As an example, the processing unit 101 can be configured to determine, based on second channel state information, that there exists another user device (AE) that satisfies the conditions that the user device can transmit model data relating to the partitioned model corresponding to the user device to the other user device via a side link corresponding to the other user device within a first predetermined time range, and that the other user device can transmit model data relating to the partitioned model corresponding to the other user device to the user device within a second predetermined time range, and then decide that the user device (UE) will participate in federative learning independently. For example, the model data relating to the partitioned model corresponding to the UE may be intermediate data #1 as described with reference to Figure 3, and the model data relating to the partitioned model corresponding to the AE may be intermediate data #2 as described with reference to Figure 3.

[0077] As an example, the processing unit 101 can be configured to determine, based on second channel state information, that there are no other user devices that satisfy the conditions that a user device can transmit model data relating to a partitioned model corresponding to another user device to another user device within a first predetermined time range, and that another user device can transmit model data relating to a partitioned model corresponding to another user device to a user device within a second predetermined time range, and then decide that the user device will independently participate in federated learning.

[0078] For example, the processing unit 101 can be configured to transmit information about the federative learning to the user device (UE) and the linked user device (AE) if the UE decides to participate in federative learning independently and selects a linked user device (AE) to train in conjunction with the user device. For example, the information about the federative learning includes the model to be trained and the division points of the model to be trained.

[0079] As an example, the processing unit 101 can be configured to broadcast the model to be trained, and the division points for dividing the model to obtain the divided models, to user devices and linked user devices. For example, the electronic device 100 broadcasts the global model (e.g., the initial global model, the updated global model) and its division points to the UE and AE in the Non-Standalone Member. In this case, both the UE and AE have the complete global model and only need to train and update the corresponding parts of the global model locally according to the division points.

[0080] For example, the processing unit 101 can be configured to send a segmented model corresponding to a user device to the user device, and a segmented model corresponding to a linked user device to the linked user device. The global model can be delivered point-to-point, meaning that the electronic device 100 can send the corresponding segmented global model to the UE and AE by individually allocating downlink resources.

[0081] As an example, the processing unit 101 is configured such that the user device (UE) and the linked user device (AE) each report their trained segmented models, and the electronic device 100 is notified to combine the trained segmented models to obtain the complete trained model. For example, referring to Figure 3, after the UE and AE in the Non-Standalone Member have completed segmented learning, the trained UE model is uploaded to the electronic device 100 by the UE, and the trained AE model is uploaded to the electronic device 100 by the AE, and they are combined in the electronic device 100 to obtain the complete trained model.

[0082] As an example, the processing unit 101 can be configured to notify the user device or a linked user device to combine the trained segmented models to obtain a complete trained model and report the complete trained model to the electronic device 100. For example, after the UE and AE in Non-Standalone Member complete segmented learning, one of the UE and AE combines the trained segmented models and uploads them to the electronic device 100. For example, the AE sends the trained AE model to the UE via sidelink, and the UE combines the trained UE model and the trained AE model before uploading the complete trained model to the electronic device 100. Alternatively, for example, the UE sends the trained UE model to the AE via sidelink, and the AE combines the trained UE model and the trained AE model before uploading the complete trained model to the electronic device 100.

[0083] For example, the processing unit 101 can be configured to notify the user device and the linked user device of information regarding side links between the user device and the linked user device.

[0084] Figure 5 shows an example of the configuration of a federated learning network according to an embodiment of the present disclosure.

[0085] Figure 5 shows an example of a federative learning network configuration, using V2X (Vehicle-to-Everything) as an example. As shown in Figure 5, in a distributed network, multiple members perform federative learning training together with electronic device 100.

[0086] In Figure 5, some Members may consist of only a single UE (for example, Member#i consists only of UE#i). Such Members are Standalone Members. In this case, the complete model to be trained is trained on the UE.

[0087] In Figure 5, some Members may consist of a single UE and a nearby AE, and such Members are Non-Standalone Members. In this case, the complete model to be trained (i.e., the complete model) is divided between the UE and AE, for example, into a UE model and an AE model. For example, Member #1 consists of UE #1 and AE #1, and the complete model to be trained is divided into the UE #1 model corresponding to UE #1 (the left-hand portion of the complete model) and the AE #1 model corresponding to AE #1 (the right-hand portion of the complete model), and the UE #1 model and AE #1 model are combined to form the complete model to be trained. For example, Member #k consists of UE #k and AE #k, and the complete model to be trained is divided into the UE #k model corresponding to UE #k (the left-hand portion of the complete model) and the AE #k model corresponding to AE #k (the right-hand portion of the complete model), and the UE #k model and AE #k model are combined to form the complete model to be trained. Also, for different Non-Standalone Members, the point of division between the UE model and the AE model may differ. For example, as can be seen by referring to Figure 4, the division point corresponding to Member#1 is at m=2, and the division point corresponding to Member#k is at m=3.

[0088] Before the start of federated learning training, devices in the network (e.g., devices that can participate in federated learning as UEs or AEs) report their self-state information to electronic device 100. For example, based on the self-state information uploaded from each device, electronic device 100 may perform at least some of the following: 1) Selecting a UE (determining whether the UE will participate in federated learning); 2) If the UE decides to participate in federated learning, deciding whether the UE will participate as a Standalone Member or a Non-Standalone Member; 3) If the UE participates in federated learning as a Non-Standalone Member, selecting an AE to work with it and determining the division points of the model to be trained; and 4) Allocating uplink transmission resources to Standalone Members, allocating uplink and sidelink transmission resources to UEs and AEs among the Non-Standalone Members, and determining how to upload the trained UE model and trained AE model to electronic device 100. During the federated learning training process, steps 1) to 4) may be repeated, i.e., UEs may be re-selected.

[0089] The following briefly describes an example of the training process of the federated learning network shown in Figure 5. 1) Electronic device 100 initializes the parameters of the global model and sends the global model. 2) Standalone Members that receive the global model perform training updates of their local models based on their local data, and after completion, upload the updated local models to electronic device 100. Non-Standalone Members perform training updates of their local models based on their UE's local data (split learning is performed between the UE and AE, and intermediate data is sent to each other via sidelinks). After completion, they upload the updated local models to electronic device 100. 3) After receiving the local models uploaded by all Members, electronic device 100 performs global model aggregation and sends the aggregated global model to each Member. 4) Steps 2) and 3) are repeated several times until the trained global model converges.

[0090] This disclosure further provides electronic equipment for wireless communication according to another embodiment. Figure 6 shows a functional block diagram of electronic equipment 600 for wireless communication according to another embodiment of this disclosure.

[0091] As shown in Figure 6, the electronic device 600 includes a communication unit 601 that can report its own state information to the network-side device providing services to the electronic device 600, so that the network-side device can determine whether the electronic device 600 will participate in federated learning in a non-independent manner by training the divided models obtained by dividing the model to be trained in cooperation with other electronic devices, or whether it will participate in federated learning independently by training the model to be trained independently.

[0092] The communication unit 601 may be implemented by one or more processing circuits, which may be implemented, for example, as a chip.

[0093] The electronic device 600 may, for example, be installed on the user equipment (UE) side or be connected to the user equipment in a communicative manner. Here, the electronic device 600 may be implemented at the chip level or at the device level. For example, the electronic device 600 may operate as the user equipment itself and may further include external devices such as memory and transceivers (not shown). The memory is used to store programs executed by the user equipment to perform various functions and related data information. The transceiver may include one or more communication interfaces to support communication between different devices (e.g., base stations, other user equipment, etc.), but the implementation of the transceiver is not specifically limited here.

[0094] For example, the network-side device may be the electronic device 100 described above. For example, the electronic device 600 may be a UE according to an embodiment of the electronic device 100 described above, and other electronic devices that cooperate with the electronic device 600 may be AEs according to an embodiment of the electronic device 100 described above.

[0095] The wireless communication system described herein may be a 5G NR communication system. Furthermore, the wireless communication system described herein may include a non-terrestrial network. Optionally, the wireless communication system described herein may also include a terrestrial network. Those skilled in the art will also understand that the wireless communication system described herein may be a 4G or 3G communication system.

[0096] For example, electronic device 600 has self-state information Info U It estimates this and reports it to the network-side device.

[0097] For explanations regarding associative learning, training models, segmented learning, etc., please refer to the explanations using Figures 2 and 3 in the embodiment of the electronic device 100, and therefore will not be repeated here.

[0098] Electronic device 600 can be a Standalone Member, in which case it does not participate in segmented learning, but independently trains a complete target model locally, and its trained local model performs global model aggregation with the trained local models of other Members to perform federative learning. Alternatively, electronic device 600 can be configured as a Non-Standalone Member with AE, perform segmented learning on the target model locally, and its trained local model performs global model aggregation with the trained local models of other Members to perform federative learning.

[0099] In embodiments of this disclosure, the electronic device 600 can perform federated learning more efficiently by reporting its own state information to the network-side device so that the network-side device can determine whether the electronic device 600 participates in federated learning independently or independently. For example, the computing and communication capabilities of different electronic devices participating in federated learning may differ; some electronic devices may have important data but lack computing or communication capabilities, while others may have strong computing and communication capabilities. For example, in the case of an electronic device with insufficient computing or communication capabilities, the load on the electronic device participating in the federated learning process can be reduced by training a segmented model in cooperation with other user devices. In the case of an electronic device with strong computing and communication capabilities, the training of the model to be trained can be completed independently locally without coordinating with other electronic devices. It can also fully utilize the computing or communication resources of other electronic devices in the communication network. Furthermore, since the number of users participating in federated learning that a network-side device can generally support is limited, it is necessary to select the electronic devices before starting federated learning. Some devices that have not been selected but wish to participate in (contribute to) federated learning can participate in training as other electronic devices. In the case of a Non-Standalone Member, data privacy and security are ensured by having only the parameters of some models of the training target model on other electronic devices, and by transmitting intermediate data between the electronic devices, thus further enhancing data privacy protection.

[0100] As an example, the self-state information of the electronic device 600 includes first channel state information of the link between the electronic device 600 and the network-side device.

[0101] As an example, the first channel status information includes at least one of the following: signal-to-interference noise ratio (SINR), reference signal received power (RSRP), reference signal received quality (RSRQ), and reference signal received strength (RSRI) of the uplink between the electronic device 600 and the network-side device.

[0102] As an example, the self-state information of the electronic device 600 includes computing and storage capability information of the electronic device 600.

[0103] As an example, computing and storage capability information includes at least one of the following: the CPU usage of the electronic device 600 and its memory size.

[0104] For example, the self-state information of the electronic device 600 includes the local data information of the electronic device 600.

[0105] As an example, local data information includes the number of training samples used when the electronic device 600 independently participates in federative learning, and the sample dimensions.

[0106] As an example, the self-state information of the electronic device 600 includes the power consumption information of the electronic device 600.

[0107] As an example, the self-state information of the electronic device 600 includes the location information of the electronic device 600.

[0108] As an example, the self-state information of the electronic device 600 includes the movement information of the electronic device 600.

[0109] As an example, the movement information includes at least one of the following: the speed of movement of the electronic device 600, the direction of movement, and the duration of stay.

[0110] As an example, the self-state information of the electronic device 600 includes second channel state information of side links between the electronic device 600 and other electronic devices within a predetermined range of the electronic device 600.

[0111] Regarding how the network-side device determines whether electronic device 600 will independently participate in federated learning or not, based on the self-state information reported by electronic device 600, please refer to the relevant explanation in the embodiment of electronic device 100, and it will not be repeated here.

[0112] As an example, when the communication unit 601 participates in federated learning independently, in each round of global training of the federated learning, it can be configured to transmit first data obtained by training the first partition model to the first partition model corresponding to the electronic device 600 to a cooperating electronic device that trains in conjunction with the electronic device 600, and to update the first partition model based on second data received from the cooperating electronic device, which is obtained by the cooperating electronic device training a second partition model corresponding to the cooperating electronic device based on the first data, by performing this a predetermined number of times to obtain the first partition model after training. As an example, the electronic device 600 is the UE described with reference to Figure 3, the first partition model is the UE model described with reference to Figure 3, the cooperating electronic device is the AE described with reference to Figure 3, the first data is intermediate data #1 described with reference to Figure 3, the second data is intermediate data #2 described with reference to Figure 3, and the second partition model is the AE model described with reference to Figure 3.

[0113] As an example, the communication unit 601 can be configured to report the first segmented model after training to the network-side device so that the network-side device can combine the first segmented model after training with the second segmented model after training received from the cooperating electronic device to obtain the complete post-trained model. The second segmented model after training is obtained when the cooperating electronic device trains the second segmented model in global training. Referring to Figure 3, after the electronic device 600 and the cooperating electronic device have completed segmented learning, the post-trained UE model is uploaded from the electronic device 600 to the network-side device, and the post-trained AE model is uploaded from the cooperating electronic device to the network-side device, where the network-side device combines them to obtain the complete post-trained model.

[0114] For example, the communication unit 601 can be configured to combine the first segmented model after training with the second segmented model after training to obtain a complete post-trained model, and to report the complete post-trained model to the network-side device. The second segmented model after training is obtained when the cooperating electronic device trains the second segmented model in global training. Referring to Figure 3, the AE transmits the AE model after training to the electronic device 600 via sidelink, and the electronic device 600 combines the UE model after training with the AE model after training, and then uploads the complete post-trained model to the network-side device.

[0115] This disclosure further provides electronic devices for wireless communication according to yet another embodiment. Figure 7 shows a functional block diagram of electronic device 700 for wireless communication according to yet another embodiment of this disclosure.

[0116] As shown in Figure 7, the electronic device 700 includes a reporting unit 701 that can report its own state information to the network-side device providing services to the electronic device 700, so that the network-side device can determine whether, in federated learning, the electronic device 700 can divide the model to be trained and train the resulting divided models in cooperation with other electronic devices, and whether the other electronic devices will participate in federated learning independently or whether the other electronic devices will train the model to be trained independently and participate in federated learning independently.

[0117] The reporting unit 701 may be implemented by one or more processing circuits, and these processing circuits may be implemented, for example, as a chip.

[0118] The electronic device 700 may, for example, be installed on the user equipment side or connected to the user equipment in a communicative manner. Here, the electronic device 700 may be implemented at the chip level or at the device level. For example, the electronic device 700 may operate as the user equipment itself, and may further include external devices such as memory and transceivers (not shown). The memory is used to store programs executed by the user equipment to perform various functions, and related data information. The transceiver may include one or more communication interfaces to support communication between different devices (e.g., base stations, other user equipment, etc.), but the implementation of the transceiver is not specifically limited here.

[0119] For example, the network-side device may be the electronic device 100 described above. For example, the electronic device 700 may be an AE according to an embodiment of the electronic device 100 described above, and other electronic devices that cooperate with the electronic device 700 may be UEs according to an embodiment of the electronic device 100 described above.

[0120] The wireless communication system described herein may be a 5G NR communication system. Furthermore, the wireless communication system described herein may include a non-terrestrial network. Optionally, the wireless communication system described herein may also include a terrestrial network. Those skilled in the art will also understand that the wireless communication system described herein may be a 4G or 3G communication system.

[0121] For example, electronic device 700 has self-state information Info A It estimates this and reports it to the network-side device.

[0122] Associative learning, the model to be trained, and segmented learning are explained in reference to Figures 2 and 3 in the embodiment of the electronic device 100, so they will not be repeated here.

[0123] Electronic device 700 forms a Non-Standalone Member with other electronic devices (UEs), performs segmented learning on the target model locally, and the trained local model of this Non-Standalone Member, together with the trained local models of the other Members, performs global model aggregation to perform federative learning.

[0124] In the embodiments of this disclosure, the electronic device 700 can perform federated learning more efficiently by reporting its own state information to the network-side device so that the network-side device can determine whether other electronic devices participate in federated learning independently or independently. For example, the computing or communication resources of the electronic device 700 can be fully utilized. Furthermore, in the case of a Non-Standalone Member in federated learning, the electronic device 700 only has parameters for some of the models of the model under training, and intermediate data is transmitted between the electronic device 700 and other electronic devices, thereby ensuring data privacy and security, i.e., further enhancing data privacy protection.

[0125] As an example, the self-state information of the electronic device 700 includes third channel state information of the link between the electronic device 700 and the network-side device.

[0126] As an example, the third channel status information includes at least one of the following: signal-to-interference noise ratio (SINR), reference signal received power (RSRP), reference signal received quality (RSRQ), and reference signal received strength (RSRI) of the uplink between the electronic device 700 and the network-side device.

[0127] As an example, the self-state information of the electronic device 700 includes computing and storage capability information of the electronic device 700.

[0128] As an example, computing and storage capability information includes at least one of the following: the CPU usage of the electronic device 700 and its memory size.

[0129] As an example, the self-state information of the electronic device 700 includes the local data information of the electronic device 700.

[0130] As an example, local data information includes the number of training samples and the sample dimensions used when the electronic device 700 independently references them for federative learning.

[0131] For example, the self-state information of the electronic device 700 includes the power consumption information of the electronic device 700.

[0132] As an example, the self-state information of the electronic device 700 includes the location information of the electronic device 700.

[0133] As an example, the self-state information of the electronic device 700 includes the movement information of the electronic device 700.

[0134] As an example, the movement information includes at least one of the following: the speed of movement of the electronic device 700, the direction of movement, and the duration of stay.

[0135] As an example, the self-state information of the electronic device 700 includes a fourth channel state information of a side link between the electronic device 700 and other electronic devices within a predetermined range of the electronic device 700.

[0136] As an example, the reporting unit 701 may be configured to obtain a trained second partition model by performing the following a predetermined number of times: when other electronic devices participate in federative learning independently, in each round of global training of the federative learning, it receives first data from other electronic devices obtained by training a first partition model corresponding to other electronic devices on a second partition model corresponding to electronic device 700, trains the second partition model based on the first data to obtain second data, and transmits the second data to the other electronic devices so that the other electronic devices update the first partition model. As an example, the other electronic devices may be UE as described with reference to Figure 3, the first partition model may be the UE model as described with reference to Figure 3, the first data may be intermediate data #1 as described with reference to Figure 3, electronic device 700 may be AE as described with reference to Figure 3, the second partition model may be the AE model as described with reference to Figure 3, and the second data may be intermediate data #2 as described with reference to Figure 3.

[0137] For example, the reporting unit 701 can be configured to report the trained second partitioned model to the network-side device so that the network-side device can combine the trained second partitioned model with the trained first partitioned model received from other electronic devices to obtain a complete trained model. The trained first partitioned model is obtained when other electronic devices train the first partitioned model in global training. Referring to Figure 3, after electronic device 700 and other electronic devices have completed partition learning, the trained AE model is uploaded from electronic device 700 to the network-side device, and the trained UE model is uploaded from other electronic devices to the network-side device, where the network-side device combines them to obtain a complete trained model.

[0138] For example, the reporting unit 701 can be configured to combine the second segmented model after training with the first segmented model after training to obtain a complete post-trained model, and to report the complete post-trained model to the network-side device. The first segmented model after training is obtained when other electronic devices train the first segmented model in global training. Referring to Figure 3, the UE transmits the UE model after training to the electronic device 700 via sidelink, and the electronic device 700 combines the UE model after training with the AE model after training, and then uploads the complete post-trained model to the network-side device.

[0139] This disclosure further provides a wireless communication system according to one embodiment, the wireless communication system comprising electronic equipment 100, electronic equipment 600, and electronic equipment 700. In this wireless communication system, referring to Figure 5, electronic equipment 100 can be a base station, electronic equipment 600 can be an UE, and electronic equipment 700 can be an AE.

[0140] In describing the electronic devices for wireless communication in the embodiments described above, several processes or methods have obviously been disclosed. The following outlines these methods without repeating some of the details already discussed in the preamble. While these methods were disclosed in the process of describing the electronic devices for wireless communication, they do not necessarily utilize or be implemented by the components described. For example, embodiments of electronic devices for wireless communication may be implemented partially or entirely by hardware and / or firmware, while the following methods for wireless communication may be implemented entirely by computer-executable programs. Of course, these methods may also utilize the hardware and / or firmware of the electronic devices for wireless communication.

[0141] Figure 8 shows a flowchart of Method S800 for wireless communication according to one embodiment of the present disclosure. Method S800 is started in step S802. In step S804, in response to self-state information reported by user devices within the service range of the electronic device, it is determined whether the user device will participate in federated learning non-independently by training the divided models obtained by dividing the model to be trained in cooperation with other user devices, or whether it will participate in federated learning independently by training the model to be trained independently. Method S800 is terminated in step S806.

[0142] This method may be performed, for example, by the electronic device 100 described above. For specific details, please refer to the above description of the related processing of the electronic device 100, but will not be repeated here.

[0143] Figure 9 shows a flowchart of Method S900 for wireless communication according to another embodiment of the present disclosure. Method S900 begins in step S902. In step S904, a network-side device providing services to an electronic device reports self-state information of the electronic device to the network-side device so that the electronic device can decide whether to participate in federated learning in a non-independent manner by training the divided models obtained by dividing the model to be trained in cooperation with other electronic devices, or to participate in federated learning independently by training the model to be trained independently. Method S900 ends in step S906.

[0144] This method may be performed, for example, by the electronic device 600 described above. For specific details, please refer to the above description of the related processing of the electronic device 600, but will not be repeated here.

[0145] Figure 10 shows a flowchart of Method S1000 for wireless communication according to another further embodiment of the present disclosure. Method S1000 begins in step S1002. In step S1004, a network-side device providing services to an electronic device reports the self-state information of the electronic device to the network-side device so that the network-side device can determine whether, in federated learning, the electronic device can divide the model to be trained and train the resulting divided models in cooperation with other electronic devices, and whether the other electronic devices participate in federated learning independently or whether the other electronic devices train the model to be trained independently and participate in federated learning independently. Method S1000 ends in step S1006.

[0146] This method may be performed, for example, by the electronic device 700 described above. For specific details, please refer to the above description of the related processing of the electronic device 700, but will not be repeated here.

[0147] The technology described in this disclosure is applicable to a variety of products.

[0148] The electronic device 100 can be implemented as various network-side devices, such as a base station. The base station can be implemented as any type of eNB (evolved Node B) or gNB (5G base station). eNBs include, for example, macro eNBs and small eNBs. Small eNBs may be eNBs that cover cells smaller than macrocells, such as pico eNBs, micro eNBs, and home (femto) eNBs. The same may apply to gNBs. Alternatively, the base station can be implemented as any other type of base station, such as a Node B or a base station transceiver (BTS). The base station can include a principal (also called base station equipment) configured to control radio communication and one or more remote radio heads (RRHs) located separately from the principal. Also, various types of user equipment can operate as a base station by temporarily or semi-permanently performing base station functions.

[0149] For example, electronic devices 600 and 700 may be implemented as various user devices. User devices may be implemented as mobile terminals (e.g., smartphones, tablet personal computers (PCs), notebook PCs, portable game consoles, portable / dongle mobile routers, and digital imaging devices) or in-vehicle terminals (e.g., car navigation systems). User devices may also be implemented as terminals that perform machine-to-machine (M2M) communication (also called machine-type communication (MTC) terminals). Furthermore, user devices may be wireless communication modules (e.g., integrated circuit modules including a single chip) mounted on each of these terminals.

[0150] [Examples of application for base stations] (First application example) Figure 11 is a block diagram showing a first example of a schematic configuration of an eNB or gNB to which the technology described herein can be applied. The following description uses an eNB as an example, but is similarly applicable to a gNB. The eNB 800 has one or more antennas 810 and a base station device 820. The base station device 820 and each antenna 810 may be connected to each other via an RF cable.

[0151] Each of the antennas 810 includes one or more antenna elements (for example, multiple antenna elements included in a MIMO antenna) and is used for transmitting and receiving radio signals by the base station equipment 820. The eNB800 may include multiple antennas 810, as shown in Figure 11. Multiple antennas 810 may be compatible with multiple frequency bands used by the eNB800, for example. Although Figure 11 shows an example in which the eNB800 includes multiple antennas 810, the eNB800 may also include a single antenna 810.

[0152] The base station device 820 includes a controller 821, a memory 822, a network interface 823, and a wireless communication interface 825.

[0153] The controller 821 may be, for example, a CPU or a DSP, and operates various functions of the upper layer of the base station equipment 820. For example, the controller 821 generates data packets from data in signals processed by the wireless communication interface 825 and forwards the generated packets via the network interface 823. The controller 821 can generate bundle packets by bundling data from multiple baseband processors and forward the generated bundle packets. The controller 821 may also have logical functions to perform controls such as radio resource control, radio bearer control, mobility management, admission control, or scheduling. Furthermore, these controls may be performed in cooperation with nearby eNBs or core network nodes. The memory 822 includes RAM and ROM and stores programs executed by the controller 821, as well as various control data (e.g., terminal lists, transmit power data, and scheduling data).

[0154] Network interface 823 is a communication interface for connecting base station equipment 820 to core network 824. Controller 821 can communicate with core network nodes or other eNBs via network interface 823. In this case, eNB 800 and the core network nodes or other eNBs are connected to each other by logical interfaces (e.g., S1 interface and X2 interface). Network interface 823 may be a wired communication interface or a wireless communication interface for a wireless backhaul line. If network interface 823 is a wireless communication interface, it can use a higher frequency band for wireless communication than the frequency band used by wireless communication interface 825.

[0155] The wireless communication interface 825 supports any cellular communication scheme (e.g., Long Term Evolution (LTE) and LTE-Advanced) and provides wireless connectivity to terminals located in the cells of the eNB800 via the antenna 810. The wireless communication interface 825 may typically include, for example, a baseband (BB) processor 826 and RF circuitry 827. The BB processor 826 can perform, for example, coding / decoding, modulation / demodulation, and multiplexing / demultiplexing, and can perform signal processing of various layers (e.g., L1, Media Access Control (MAC), Radio Link Control (RLC), Packet Data Aggregation Protocol (PDCP)). The BB processor 826 may have some or all of the above logical functions instead of the controller 821. The BB processor 826 may be a memory that stores a communication control program, or it may be a module that includes a processor and associated circuitry configured to execute the program. Program updates can change the functionality of the BB processor 826. This module may be a card or blade inserted into a slot in the base station equipment 820. Alternatively, this module may be a chip mounted on a card or blade. At the same time, the RF circuit 827 may include, for example, a mixer, a filter, and an amplifier, and may transmit and receive radio signals via the antenna 810.

[0156] As shown in Figure 11, the wireless communication interface 825 may include multiple BB processors 826. For example, multiple BB processors 826 may be compatible with multiple frequency bands used by the eNB800. As shown in Figure 11, the wireless communication interface 825 may include multiple RF circuits 827. For example, multiple RF circuits 827 may be compatible with multiple antenna elements. Although Figure 11 shows an example in which the wireless communication interface 825 includes multiple BB processors 826 and multiple RF circuits 827, the wireless communication interface 825 may include a single BB processor 826 or a single RF circuit 827.

[0157] In the eNB800 shown in Figure 11, when the electronic device 100 is implemented as a base station, the transceiver may be implemented by the wireless communication interface 825. At least part of the function may be implemented by the controller 821. For example, the controller 821 can perform federated learning more efficiently by executing the functions of the electronic device 100.

[0158] (Second application example) Figure 12 is a block diagram showing a second example of a schematic configuration of an eNB or gNB to which the technology described herein can be applied. Similarly, although the following description uses an eNB as an example, it is also applicable to a 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 also be connected to each other via a high-speed line such as an optical fiber cable.

[0159] Each of the antennas 840 includes one or more antenna elements (for example, multiple antenna elements included in a MIMO antenna) and is used for transmitting and receiving radio signals by the RRH860. The eNB830 may include multiple antennas 840, as shown in Figure 12. Multiple antennas 840 may be compatible with multiple frequency bands used by the eNB830, for example. Although Figure 12 shows an example in which the eNB830 includes multiple antennas 840, the eNB830 may also include a single antenna 840.

[0160] 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, memory 852, and network interface 853 are similar to the controller 821, memory 822, and network interface 823 described with reference to Figure 13.

[0161] The wireless communication interface 855 supports any cellular communication method (e.g., LTE and LTE-Advanced) and provides wireless connectivity to terminals located in the sector corresponding to the RRH860 via the RRH860 and antenna 840. The wireless communication interface 855 may typically include, for example, a BB processor 856. The BB processor 856 is similar to the BB processor 826 described with reference to Figure 11, except that it is connected to the RF circuit 864 of the RRH860 via a connection interface 857. The wireless communication interface 855 may include multiple BB processors 856, as shown in Figure 12. Multiple BB processors 856 may be compatible with multiple frequency bands used by, for example, the eNB830. Although Figure 12 shows an example in which the wireless communication interface 855 includes multiple BB processors 856, the wireless communication interface 855 may include a single BB processor 856.

[0162] The connection interface 857 is an interface for connecting the base station device 850 (wireless communication interface 855) to the RRH860. The connection interface 857 may also be a communication module for communication on the high-speed line described above for connecting the base station device 850 (wireless communication interface 855) to the RRH860.

[0163] The RRH860 includes a connection interface 861 and a wireless communication interface 863.

[0164] The connection interface 861 is an interface for connecting the RRH860 (wireless communication interface 863) to the base station device 850. The connection interface 861 may also be a communication module for communication on the high-speed line described above.

[0165] The wireless communication interface 863 transmits and receives radio signals via the antenna 840. The wireless communication interface 863 may typically include, for example, an RF circuit 864. The RF circuit 864 may include, for example, a mixer, a filter, and an amplifier, and may transmit and receive radio signals via the antenna 840. The wireless communication interface 863 may include multiple RF circuits 864, as shown in Figure 14. Multiple RF circuits 864 can support multiple antenna elements. Although Figure 12 shows an example in which the wireless communication interface 863 includes multiple RF circuits 864, the wireless communication interface 863 may include a single RF circuit 864.

[0166] In the eNB830 shown in Figure 12, when the electronic device 100 is implemented as a base station, the transceiver may be implemented by the wireless communication interface 855. At least part of the function may be implemented by the controller 851. For example, the controller 851 can perform federated learning more efficiently by executing a part of the function in the electronic device 100.

[0167] [Examples of application for user equipment] (First application example) Figure 13 is a block diagram showing an example of a schematic configuration of a smartphone 900 to which the technology described herein can be applied. The smartphone 900 includes a processor 901, memory 902, storage device 903, external connection interface 904, imaging device 906, sensor 907, microphone 908, input device 909, display device 910, speaker 911, wireless communication interface 912, one or more antenna switches 915, one or more antennas 916, bus 917, battery 918, and auxiliary controller 919.

[0168] The processor 901 is, for example, a CPU or a system-on-a-chip (SoC) and can control the functions of the application layer and other layers of the smartphone 900. The memory 902 includes RAM and ROM and stores data and programs executed by the processor 901. The storage device 903 may include, for example, semiconductor memory and storage media such as a hard disk. The external connection interface 904 is an interface for connecting external devices (for example, a memory card and a Universal Serial Bus (USB) device) to the smartphone 900.

[0169] The imaging device 906 includes an image sensor (e.g., a charge-coupled device (CCD) and a complementary metal-oxide-semiconductor (CMOS)) and generates an image. Sensor 907 may include a set of sensors such as a measuring sensor, a gyroscope, a geomagnetic sensor, and an accelerometer. Microphone 908 converts sound input to the smartphone 900 into an audio signal. Input device 909 includes, for example, a touch sensor, a keypad, a keyboard, a button, or a switch configured to detect touches on the screen of the display device 910 and receives operations or information input from the user. Display device 910 includes a screen (e.g., a liquid crystal display (LCD), an organic light-emitting diode (OLED) display) and displays the output image from the smartphone 900. Speaker 911 converts the audio signal output from the smartphone 900 into sound.

[0170] The wireless communication interface 912 supports any cellular communication method (e.g., LTE and LTE-Advanced) and performs wireless communication. The wireless communication interface 912 typically includes, for example, a broadband processor 913 and an RF circuit 914. The broadband processor 913 can perform various types of signal processing for wireless communication, such as encoding / decoding, modulation / demodulation, and multiplexing / demultiplexing. Simultaneously, the RF circuit 914 includes, for example, a mixer, filter, and amplifier, and can transmit and receive wireless signals via the antenna 916. Note that the figure shows a case where one RF link is connected to one antenna; this is merely an example, and a single RF link can also be connected to multiple antennas via multiple phase shifters. The wireless communication interface 912 can be a single chip module on which the broadband processor 913 and RF circuit 914 are integrated. As shown in Figure 13, the wireless communication interface 912 can include multiple broadband processors 913 and multiple RF circuits 914. Figure 13 shows an example in which the wireless communication interface 912 includes multiple BB processors 913 and multiple RF circuits 914, but the wireless communication interface 912 may also include a single BB processor 913 or a single RF circuit 914.

[0171] In addition to cellular communication, the wireless communication interface 912 can support other types of wireless communication, such as short-range wireless communication, proximity communication, and wireless local network (LAN) communication. In this case, the wireless communication interface 912 may include a BB processor 913 and an RF circuit 914 for various wireless communication methods.

[0172] Each of the antenna switches 915 switches the destination of the antenna 916 among multiple circuits included in the wireless communication interface 912 (for example, circuits used for different wireless communication methods).

[0173] Each of the antennas 916 includes one or more antenna elements (for example, multiple antenna elements included in a MIMO antenna) and is used to transmit and receive radio signals via the wireless communication interface 912. As shown in Figure 13, the smartphone 900 may include multiple antennas 916. Although Figure 13 shows an example in which the smartphone 900 includes multiple antennas 916, the smartphone 900 may also include a single antenna 916.

[0174] The smartphone 900 may include an antenna 916 for various wireless communication methods. In this case, the antenna switch 915 can be omitted from the configuration of the smartphone 900.

[0175] Bus 917 connects the processor 901, memory 902, storage device 903, external connection interface 904, imaging device 906, sensor 907, microphone 908, input device 909, display device 910, speaker 911, wireless communication interface 912, and auxiliary controller 919 to each other. Battery 918 supplies power to each block of the smartphone 900 shown in Figure 13 via power lines, which are represented as partially dotted lines in the drawing. The auxiliary controller 919 operates the minimum necessary functions of the smartphone 900, for example, in sleep mode.

[0176] In the smartphone 900 shown in Figure 13, if the electronic devices 600 and 700 are implemented as, for example, user-side smartphones, the transceivers of the electronic devices 600 and 700 may be implemented by the wireless communication interface 912. At least part of the functions may be implemented by the processor 901 or the auxiliary controller 919. For example, the processor 901 or the auxiliary controller 919 can perform federated learning more efficiently by executing the functions of the electronic devices 600 and 700.

[0177] (Second application example) Figure 14 is a block diagram showing an example of a schematic configuration of a car navigation device 920 to which the technology described herein can be applied. The car navigation device 920 includes a processor 921, memory 922, global positioning system (GPS) module 924, sensor 925, data interface 926, content player 927, storage medium interface 928, input device 929, display device 930, speaker 931, wireless communication interface 933, one or more antenna switches 936, one or more antennas 937, and a battery 938.

[0178] The processor 921 is, for example, a CPU or SoC, and can control the navigation and other functions of the car navigation device 920. The memory 922 includes RAM and ROM and stores data and programs executed by the processor 921.

[0179] The GPS module 924 measures the position (e.g., latitude, longitude, altitude) of the car navigation device 920 using GPS signals received from GPS satellites. The sensor 925 may include a set of sensors, such as a gyro sensor, a geomagnetic sensor, and a barometric pressure sensor. The data interface 926 is connected to, for example, an in-vehicle network 941 via a terminal (not shown) to acquire data generated by the vehicle (e.g., vehicle speed data).

[0180] The content player 927 plays content stored on a storage medium (e.g., CD and DVD) inserted into the storage medium interface 928. The input device 929 includes, for example, a touch sensor, button, or switch 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, for example, an LCD or OLED display screen, and displays images of the navigation function or the played content. The speaker 931 outputs sounds of the navigation function or the played content.

[0181] The wireless communication interface 933 supports any cellular communication scheme (e.g., LTE and LTE-Advanced) and performs wireless communication. The wireless communication interface 933 typically includes, for example, a broadband processor 934 and an RF circuit 935. The broadband processor 934 can perform various types of signal processing for wireless communication, such as coding / decoding, modulation / demodulation, and multiplexing / demultiplexing. Simultaneously, the RF circuit 935 includes, for example, a mixer, filter, and amplifier, and can transmit and receive wireless signals via the antenna 937. The wireless communication interface 933 can also be a single chip module with the broadband processor 934 and RF circuit 935 integrated on it. As shown in Figure 14, the wireless communication interface 933 can include multiple broadband processors 934 and multiple RF circuits 935. While Figure 14 shows an example where the wireless communication interface 933 includes multiple broadband processors 934 and multiple RF circuits 935, the wireless communication interface 933 may include a single broadband processor 934 or a single RF circuit 935.

[0182] In addition to cellular communication, the wireless communication interface 933 can support other types of wireless communication, such as short-range wireless communication, proximity communication, and wireless LAN. In this case, the wireless communication interface 933 can include a BB processor 934 and an RF circuit 935 for each type of wireless communication.

[0183] Each of the antenna switches 936 switches the destination of the antenna 937 among multiple circuits included in the wireless communication interface 933 (for example, circuits used for different wireless communication methods).

[0184] Each of the antennas 937 includes one or more antenna elements (for example, multiple antenna elements included in a MIMO antenna) and is used to transmit and receive radio signals via the wireless communication interface 933. As shown in Figure 14, the car navigation device 920 may include multiple antennas 937. Although Figure 14 shows an example in which the car navigation device 920 includes multiple antennas 937, the car navigation device 920 may also include a single antenna 937.

[0185] The car navigation system 920 may include an antenna 937 for various wireless communication methods. In this case, the antenna switch 936 can be omitted from the configuration of the car navigation system 920.

[0186] Battery 938 supplies power to each block of the car navigation system 920 shown in Figure 14 via power lines, which are partially represented as dotted lines in the drawing. Battery 938 stores power supplied from the vehicle.

[0187] In the car navigation device 920 shown in Figure 14, if the electronic devices 600 and 700 are implemented, for example, as car navigation devices on the user's device side, the transceivers of the electronic devices 600 and 700 may be implemented by the wireless communication interface 933. At least part of the functions may be implemented by the processor 921. For example, the processor 921 can perform associative learning more efficiently by executing the functions of the electronic devices 600 and 700.

[0188] The technology described herein may be implemented as an in-vehicle system (or vehicle) 940 including one or more blocks of a car navigation device 920, an in-vehicle network 941, and a vehicle module 942. The vehicle module 942 generates vehicle data (e.g., vehicle speed, engine speed, fault information) and outputs the generated data to the in-vehicle network 941.

[0189] The above describes the basic principles of the present invention by combining specific embodiments. However, those skilled in the art will understand that all or any step or component of the method and apparatus of the present invention can be implemented in any computer device (including processors, storage media, etc.) or network of computer devices by hardware, firmware, software, or a combination thereof, and that this can be implemented by those skilled in the art by reading the description of the present invention and using their basic circuit design knowledge or basic programming skills.

[0190] Furthermore, the present invention provides a program product that stores machine-readable instruction codes. When these instruction codes are read and executed by a device, the method according to the above-described embodiment of the present invention is performed.

[0191] Accordingly, the disclosure of the present invention also includes a storage medium for storing a program product in which the above-mentioned machine-readable instruction code is stored. The above-mentioned storage medium includes, but is not limited to, flexible disks, optical disks, magneto-optical disks, memory cards, memory sticks, etc.

[0192] When this disclosure is implemented by software or firmware, the programs constituting the software are installed from a storage medium or network to a computer having a dedicated hardware configuration (for example, the general-purpose computer 1500 shown in Figure 15), and once the various programs are installed, the computer can perform various functions.

[0193] In Figure 15, the central processing unit (CPU) 1501 executes various processes based on programs stored in the read-only memory (ROM) 1502, or programs loaded from the storage unit 1508 into the random access memory (RAM) 1503. The RAM 1503 stores data necessary for the CPU 1501 to execute various processes as needed. The CPU 1501, ROM 1502, and RAM 1503 are connected to each other via the bus 1504. The input / output interface 1505 is also connected to the bus 1504.

[0194] The input section 1506 (including keyboard, mouse, etc.), output section 1507 (including displays such as cathode ray tubes (CRTs), liquid crystal displays (LCDs), etc., and speakers, etc.), storage section 1508 (including hard disks, etc.), and communication section 1509 (including network interface cards such as LAN cards and modulators / demodulators) are connected to the input / output interface 1505. The communication section 1509 performs communication processing over a network, such as the Internet. If necessary, the drive 1510 may also be connected to the input / output interface 1505. Removable media 1511, such as magnetic disks, optical disks, magneto-optical disks, and semiconductor memory, can be mounted on the drive 1510 if necessary, so that computer programs read from them can be installed in the storage section 1508 if necessary.

[0195] When the above series of processes are implemented using software, the programs that make up the software are installed from a network such as the internet, or from a storage medium such as removable media 1511.

[0196] Those skilled in the art should understand that such a storage medium is not limited to the removable media 1511 shown in Figure 15, which stores the program and provides the program to the user by being distributed separately from the device. Examples of removable media 1511 include magnetic disks (including Flexible Disks®), optical disks (including Optical Disk Read-Only Memory (CD-ROM) and Digital General Purpose Disks (DVD)), magneto-optical disks (including MiniDisc (MD)®), and semiconductor memory. Alternatively, the storage medium may be a ROM 1502, a hard disk included in the storage section 1508, etc., which stores the program and is distributed to the user together with the device containing them.

[0197] In the apparatus, method, and system of the present invention, each component or step can be disassembled and / or reassembled. These disassembly and / or reassembly should also be considered equivalent solutions of the present invention. The execution steps of the above series of processes can be performed in the order described and in chronological order, but are not necessarily required to be performed in chronological order. Some steps may be performed in parallel or independently of each other.

[0198] Finally, the terms “include,” “incorporate,” or any other variation thereof are intended to include non-exclusive inclusion, thereby including not only those elements but also other elements not explicitly listed, or the inherent elements of such process, method, product, or device. Furthermore, unless otherwise specified, the elements limited by the phrase “include one…” do not preclude the presence of other identical elements in the process, method, product, or device that includes the aforementioned elements.

[0199] Although embodiments of the present invention have been described in detail above with reference to the drawings, it should be understood that the embodiments described above are for illustrative purposes only and do not limit the present invention. Those skilled in the art will know that various modifications and changes can be made to the above embodiments without departing from the substance and scope of the present invention. Therefore, the scope of the present invention is limited only to the appended claims and their equivalents.

[0200] This technology can be further realized as follows: Plan 1. Electronic equipment for wireless communication, An electronic device including a processing circuit configured to determine, in federated learning, whether a user device participates in the federated learning non-independently by training a divided model obtained by dividing the model to be trained in cooperation with other user devices, or participates in the federated learning independently by training the model to be trained independently, based on self-state information reported by a user device within the service scope of the electronic device. Plan 2. The self-state information includes first channel state information of the link between the user device and the electronic device, The electronic device according to Plan 1, wherein the processing circuit is configured to determine whether the user device will participate in the federated learning independently if the value of the first channel state information is smaller than a first predetermined threshold. Plan 3. The electronic device according to Plan 2, wherein the first channel state information includes at least one of the signal-to-interference noise ratio (SINR), reference signal received power (RSRP), reference signal received quality (RSRQ), and reference signal received strength (RSRI) of the uplink between the user device and the electronic device. Plan 4. The self-state information includes computing and storage capability information of the user device, The electronic device according to Solution 1, wherein the processing circuit is configured to determine whether the user device will participate in the federative learning independently if the size of the storage space indicated by the computing and storage capacity information is smaller than the size of the model to be trained, and / or the computing power indicated by the computing and storage capacity information is smaller than the computing power required to train the model to be trained. Plan 5. The electronic device according to plan 4, wherein the computing and storage capability information includes at least one of the CPU usage rate and memory size of the user device. Plan 6. The self-state information includes the local data information of the user device, The electronic device according to Solution 1, wherein the processing circuit is configured to determine whether the user device will participate in the federative learning independently if the training time required for the user device to train the model to be trained, obtained based on the local data information and the computing and storage capability information of the user device, is greater than a second predetermined threshold. Plan 7. The electronic device according to Plan 6, wherein the local data information includes the number of training samples and the sample dimensions used when the user device independently participates in the federated learning. Plan 8. The self-state information includes the power consumption information of the user equipment, The electronic device according to Plan 1, wherein the processing circuit is configured to determine, if the amount of power indicated by the power information is less than a third predetermined threshold, the user device will independently participate in the federated learning. Plan 9. The self-state information includes the location information of the user device, The electronic device according to Solution 1, wherein the processing circuit is configured to determine, based on the location information, that no other user devices exist within a predetermined range of the user device, and the user device independently decides to participate in the federated learning. Plan 10. The self-state information includes the movement information of the user device, The electronic device according to Plan 1, wherein the processing circuit is configured to determine, based on the movement information, that the user device has high mobility, and to decide that the user device independently participates in the federated learning. Plan 11. The electronic device according to plan 10, wherein the movement information includes at least one of the movement speed, movement direction, and stay time of the user device. Plan 12. The electronic device according to any one of the solutions 1 to 11, wherein the processing circuit is configured to select a collaborating user device to train in conjunction with the user device based on second channel state information of sidelinks between the user device and other user devices within a predetermined range of the user device, when the user device decides to participate in the collaborative learning independently. Plan 13. The electronic device according to solution 12, wherein the processing circuit is configured to select the other user device as the linked user device if the value of the second channel state information of the side link corresponding to the other user device is greater than a fourth predetermined threshold. Plan 14. The electronic device according to plan 13, wherein the second channel state information includes at least one of the sidelink signal-to-interference noise ratio (SINR), reference signal received power (RSRP), reference signal received quality (RSRQ), and reference signal received strength (RSRI). Plan 15. The electronic device according to solution 12, wherein the processing circuit is configured to select as the linked user device a user device that satisfies the conditions that the user device can transmit model data relating to a segmented model corresponding to the user device to the other user device within a first predetermined time range, and / or that the other user device can transmit model data relating to a segmented model corresponding to the other user device to the user device within a second predetermined time range, via a side link corresponding to the other user device. Plan 16. The electronic device according to any one of the solutions 12 to 15, wherein the processing circuit is configured to determine division points for dividing the training target model to obtain the divided model based on the second channel state information when the linked user device is selected. Plan 17. The electronic device according to solution 16, wherein the processing circuit is configured such that the determined division point satisfies the condition that the user device can transmit model data relating to a division model corresponding to the user device to the linked user device within a first predetermined time range, and / or that the linked user device can transmit model data relating to a division model corresponding to the linked user device to the user device within a second predetermined time range. Plan 18. The electronic device according to Plan 1, wherein the self-state information includes second channel state information of a side link between the user device and other user devices within a predetermined range of the user device. Plan 19. The electronic device according to plan 18, wherein the processing circuit is configured to determine, if other user devices exist within the predetermined range where the value of the second channel state information is greater than a fifth predetermined threshold, the user device will independently participate in the federated learning. Plan 20. The electronic device according to plan 19, wherein the processing circuit is configured to determine that the user device independently participates in the federated learning if the second channel state information corresponding to other user devices within the predetermined range is all below the fifth predetermined threshold. Plan 21. The electronic device according to plan 19 or 20, wherein the second channel state information includes at least one of the sidelink signal-to-interference noise ratio (SINR), reference signal received power (RSRP), reference signal received quality (RSRQ), and reference signal received strength (RSRI). Plan 22. The electronic device according to solution 18, wherein the processing circuit is configured to determine, based on the second channel state information, that there exists another user device that satisfies the conditions that the user device can transmit model data relating to a partitioned model corresponding to the user device to the other user device via a side link corresponding to the other user device within a first predetermined time range, and that the other user device can transmit model data relating to a partitioned model corresponding to the other user device to the user device within a second predetermined time range, and then decides that the user device will participate in the federated learning independently. Plan 23. The electronic device according to solution 18, wherein the processing circuit is configured to determine, based on the second channel state information, that there are no other user devices that satisfy the conditions that the user device can transmit model data relating to a partitioned model corresponding to the user device to the other user device within a first predetermined time range, and that the other user device can transmit model data relating to a partitioned model corresponding to the other user device to the user device within a second predetermined time range, and then the user device decides to independently participate in the federated learning. Plan 24. The electronic device according to any one of the solutions 1 to 23, wherein the processing circuit is configured to transmit information regarding the federative learning to the user device and the linked user device when the user device decides to participate in the federative learning independently and selects a linked user device to be trained in conjunction with the user device. Plan 25. The electronic device according to plan 24, wherein the processing circuit is configured to broadcast the training target model and the division points for dividing the training target model to obtain the divided models to the user device and the linked user device. Plan 26. The electronic device according to solution 24, wherein the processing circuit is configured to transmit a segmented model corresponding to the user device to the user device and transmit a segmented model corresponding to the linked user device to the linked user device. Plan 27. The electronic device according to any one of the solutions 24 to 26, wherein the processing circuit is configured such that the user device and the linked user device each report a segmented model after training, and the electronic device notifies the user device to combine the segmented models after training to obtain a complete model after training. Plan 28. The electronic device according to any one of the solutions 24 to 26, wherein the processing circuit is configured to notify the user device or the linked user device to combine the trained segmented models to obtain a complete trained model and report the complete trained model to the electronic device. Plan 29. The electronic device according to any one of the solutions 24 to 28, wherein the processing circuit is configured to notify the user device and the linked user device of information regarding a side link between the user device and the linked user device. Plan 30. Electronic equipment for wireless communication, An electronic device including a processing circuit configured to report the self-state information of the electronic device to the network-side device, which provides services to the electronic device, so that the network-side device determines whether the electronic device will participate in the federative learning in a non-independent manner by training the divided models obtained by dividing the model to be trained in cooperation with other electronic devices, or whether it will participate in the federative learning independently by training the model to be trained independently. Plan 31. The electronic device according to the solution 30, wherein the self-state information includes first channel state information of the link between the electronic device and the network-side device. Plan 32. The electronic device according to plan 31, wherein the first channel state information includes at least one of the signal-to-interference noise ratio (SINR), reference signal received power (RSRP), reference signal received quality (RSRQ), and reference signal received strength (RSRI) of the uplink between the electronic device and the network-side device. Plan 33. The electronic device according to plan 30, wherein the self-state information includes computing and storage capability information of the electronic device. Plan 34. The electronic device according to plan 33, wherein the computing and storage capability information includes at least one of the CPU usage rate and memory size of the electronic device. Plan 35. The electronic device according to plan 30, wherein the self-state information includes local data information of the electronic device. Plan 36. The electronic device according to plan 35, wherein the local data information includes the number of training samples and the sample dimensions used when the electronic device independently participates in the federated learning. Plan 37. The electronic device described in plan 30, wherein the self-state information includes the power consumption information of the electronic device. Plan 38. The electronic device according to plan 30, wherein the self-state information includes the location information of the electronic device. Plan 39. The electronic device according to plan 30, wherein the self-state information includes the movement information of the electronic device. Plan 40. The electronic device according to plan 39, wherein the movement information includes at least one of the movement speed, movement direction, and stay time of the electronic device. Plan 41. The electronic device according to plan 30, wherein the self-state information includes second channel state information of a side link between the electronic device and other electronic devices within a predetermined range of the electronic device. Plan 42. The aforementioned processing circuit is When participating in the aforementioned federative learning independently, in each round of the global training of the aforementioned federative learning, with respect to the first segmented model corresponding to the electronic device, The first data obtained by training the first segmented model is transmitted to a linked electronic device that trains in conjunction with the electronic device. The linked electronic device updates the first partition model based on the second data received from the linked electronic device, which is obtained by training a second partition model corresponding to the linked electronic device based on the first data. An electronic device according to any one of the solutions 30 to 41, configured to obtain a first split model after training by performing the following a predetermined number of times. Plan 43. The processing circuit is configured to report the first segmented model after training to the network-side device so that the network-side device can combine the first segmented model after training with the second segmented model after training received from the linked electronic device to obtain a complete model after training. The electronic device according to plan 42, wherein the second segmented model after the training is obtained by the collaborative electronic device training the second segmented model in the global training. Plan 44. The processing circuit is configured to combine the first partitioned model after training and the second partitioned model after training to obtain a complete model after training, and to report the complete model after training to the network-side device. The electronic device according to plan 42, wherein the second segmented model after the training is obtained by the collaborative electronic device training the second segmented model in the global training. Plan 45. Electronic equipment for wireless communication, An electronic device including a processing circuit configured to report self-state information of the electronic device to the network-side device providing services to the electronic device, so as to determine whether the electronic device can train the divided models obtained by dividing the model to be trained in cooperation with other electronic devices during federated learning, and whether the other electronic devices participate in the federated learning independently or whether the other electronic devices train the model to be trained independently and participate in the federated learning independently. Plan 46. The electronic device according to plan 45, wherein the self-state information includes third channel state information of the link between the electronic device and the network-side device. Plan 47. The electronic device according to plan 46, wherein the third channel state information includes at least one of the signal-to-interference noise ratio (SINR), reference signal received power (RSRP), reference signal received quality (RSRQ), and reference signal received strength (RSRI) of the uplink between the electronic device and the network-side device. Plan 48. The electronic device according to plan 45, wherein the self-state information includes computing and storage capability information of the electronic device. Plan 49. The electronic device according to plan 48, wherein the computing and storage capability information includes at least one of the CPU usage rate and memory size of the electronic device. Plan 50. The electronic device according to plan 49, wherein the self-state information includes local data information of the electronic device. Plan 51. The electronic device according to Scheme 50, wherein the local data information includes the number of training samples and the sample dimensions used when the electronic device independently participates in the federated learning. Plan 52. The electronic device described in plan 45, wherein the self-state information includes the power consumption information of the electronic device. Plan 53. The electronic device according to plan 45, wherein the self-state information includes the location information of the electronic device. Plan 54. The electronic device according to plan 45, wherein the self-state information includes the movement information of the electronic device. Plan 55. The electronic device according to plan 54, wherein the movement information includes at least one of the movement speed, movement direction, and stay time of the electronic device. Plan 56. The electronic device according to plan 45, wherein the self-state information includes a fourth channel state information of a side link between the electronic device and other electronic devices within a predetermined range of the electronic device. Plan 57. The aforementioned processing circuit is If the other electronic devices participate in the federative learning independently, in each round of the global training of the federative learning, with respect to the second partition model corresponding to the electronic device, From the aforementioned other electronic device, first data obtained by training a first segmentation model corresponding to the aforementioned other electronic device is received. The second partition model is trained based on the first data to obtain second data, and the second data is transmitted to the other electronic device so that the other electronic device updates the first partition model. An electronic device according to any one of the solutions 45 to 56, configured to obtain a second split model after training by performing the following a predetermined number of times. Plan 58. The processing circuit is configured to report the trained second segmented model to the network-side device so that the network-side device can combine the trained second segmented model with the trained first segmented model received from the other electronic device to obtain a complete trained model. The electronic device according to plan 57, wherein the first segmented model after training is obtained by training the first segmented model with the other electronic device in the global training. Plan 59. The processing circuit is configured to combine the second partitioned model after training with the first partitioned model after training to obtain a complete model after training, and to report the complete model after training to the network-side device. The electronic device according to plan 57, wherein the first segmented model after training is obtained by training the first segmented model with the other electronic device in the global training. Plan 60. An electronic device described in any one of the solutions 1 to 29, An electronic device described in any one of the plans 30 to 44, A wireless communication system including an electronic device as described in any one of the solutions 45 to 59. Plan 61. A method for wireless communication, A method comprising determining, in federated learning, whether a user device participates in the federated learning non-independently by training a divided model obtained by dividing the model to be trained in cooperation with other user devices, or independently by training the model to be trained independently, based on self-state information reported by a user device within the service scope of an electronic device. Plan 62. A method for wireless communication, A method comprising reporting the self-state information of an electronic device to the network-side device providing services to an electronic device, so that the electronic device can determine whether to participate in the federative learning in a non-independent manner by training the divided models obtained by dividing the model to be trained in cooperation with other electronic devices, or to participate in the federative learning independently by training the model to be trained independently. Plan 63. A method for wireless communication, A method comprising a network-side device providing services to an electronic device reporting the self-state information of the electronic device to the network-side device to determine whether the electronic device can train the partitioned models obtained by partitioning the model to be trained in cooperation with other electronic devices in federated learning, and whether the other electronic devices participate in the federated learning independently or whether the other electronic devices train the model to be trained independently and participate in the federated learning independently. Plan 64. A computer-readable storage medium that stores computer-executable instructions causing, if executed, to perform a method for wireless communication described in any one of the schemes 61 to 63.

Claims

1. Electronic equipment for wireless communication, An electronic device including a processing circuit configured to determine, in federated learning, whether a user device participates in the federated learning non-independently by training a divided model obtained by dividing the model to be trained in cooperation with other user devices, or participates in the federated learning independently by training the model to be trained independently, based on self-state information reported by a user device within the service scope of the electronic device.

2. The self-state information includes first channel state information of the link between the user device and the electronic device, The electronic device according to claim 1, wherein the processing circuit is configured to determine whether the user device will participate in the federated learning independently if the value of the first channel state information is smaller than a first predetermined threshold.

3. The electronic device according to claim 2, wherein the first channel state information includes at least one of the signal-to-interference noise ratio SINR of the uplink between the user device and the electronic device, reference signal received power RSRP, reference signal received quality RSRQ, and reference signal received intensity RSRI.

4. The self-state information includes computing and storage capability information of the user device, The electronic device according to claim 1, wherein the processing circuit is configured to determine, independently of the user device, whether the size of the storage space indicated by the computing and storage capacity information is smaller than the size of the model to be trained, and / or the computing capacity indicated by the computing and storage capacity information is smaller than the computing capacity required to train the model to be trained.

5. The electronic device according to claim 4, wherein the computing and storage capability information includes at least one of the CPU usage rate and memory size of the user device.

6. The self-state information includes the local data information of the user device, The electronic device according to claim 1, wherein the processing circuit is configured to determine whether the user device will participate in the federative learning independently if the training time required for the user device to train the model to be trained, obtained based on the local data information and the computing and storage capability information of the user device, is greater than a second predetermined threshold.

7. The electronic device according to claim 6, wherein the local data information includes the number of training samples and the sample dimension used when the user device independently participates in the federated learning.

8. The self-state information includes the power consumption information of the user equipment, The electronic device according to claim 1, wherein the processing circuit is configured to determine, when the amount of power indicated by the power information is less than a third predetermined threshold, that the user device will participate in the federated learning independently.

9. The self-state information includes the location information of the user device, The electronic device according to claim 1, wherein the processing circuit is configured to determine, based on the location information, that no other user devices exist within a predetermined range of the user device, and to decide that the user device independently participates in the federated learning.

10. The self-state information includes the movement information of the user device, The electronic device according to claim 1, wherein the processing circuit is configured to determine, based on the movement information, that the user device has high mobility, and to decide that the user device independently participates in the federated learning.

11. The electronic device according to claim 10, wherein the movement information includes at least one of the movement speed, movement direction, and stay time of the user device.

12. The electronic device according to any one of claims 1 to 11, wherein the processing circuit is configured to select a collaborating user device to train in conjunction with the user device based on second channel state information of sidelinks between the user device and other user devices within a predetermined range of the user device when the user device decides to participate in the collaborative learning independently.

13. The electronic device according to claim 12, wherein the processing circuit is configured to select the other user device as the linked user device if the value of the second channel state information of the side link corresponding to the other user device is greater than a fourth predetermined threshold.

14. The electronic device according to claim 13, wherein the second channel state information includes at least one of the sidelink signal-to-interference noise ratio SINR, reference signal received power RSRP, reference signal received quality RSRQ, and reference signal received intensity RSRI.

15. The electronic device according to claim 12, wherein the processing circuit is configured to select, based on the second channel state information, an other user device that satisfies the conditions that the user device can transmit model data relating to a segmented model corresponding to the user device to the other user device within a first predetermined time range, and / or that the other user device can transmit model data relating to a segmented model corresponding to the other user device to the user device within a second predetermined time range, via a side link corresponding to the other user device, as the linked user device.

16. The electronic device according to any one of claims 12 to 15, wherein the processing circuit is configured to determine division points for dividing the training target model to obtain the divided model based on the second channel state information when the linked user device is selected.

17. The electronic device according to claim 16, wherein the processing circuit is configured such that the determined division point satisfies the condition that the user device can transmit model data relating to a division model corresponding to the user device to the linked user device within a first predetermined time range, and / or that the linked user device can transmit model data relating to a division model corresponding to the linked user device to the user device within a second predetermined time range.

18. The electronic device according to claim 1, wherein the self-state information includes second channel state information of a side link between the user device and other user devices within a predetermined range of the user device.

19. The electronic device according to claim 18, wherein the processing circuit is configured to determine, if there are other user devices within the predetermined range whose second channel state information value is greater than a fifth predetermined threshold, the user device will participate in the federated learning independently.

20. The electronic device according to claim 19, wherein the processing circuit is configured to determine that the user device independently participates in the federated learning if all second channel state information corresponding to other user devices within the predetermined range is below the fifth predetermined threshold.

21. The electronic device according to claim 19 or 20, wherein the second channel state information includes at least one of the sidelink signal-to-interference noise ratio SINR, reference signal received power RSRP, reference signal received quality RSRQ, and reference signal received intensity RSRI.

22. The electronic device according to claim 18, wherein the processing circuit is configured to determine, based on the second channel state information, that there exists another user device that satisfies the conditions that the user device can transmit model data relating to a partitioned model corresponding to the user device to the other user device via a side link corresponding to the other user device within a first predetermined time range, and that the other user device can transmit model data relating to a partitioned model corresponding to the other user device to the user device within a second predetermined time range, and then decides that the user device will participate in the federated learning independently.

23. The electronic device according to claim 18, wherein the processing circuit is configured to determine, based on the second channel state information, that there are no other user devices that satisfy the conditions that the user device can transmit model data relating to a partitioned model corresponding to the user device to the other user device within a first predetermined time range, and that the other user device can transmit model data relating to a partitioned model corresponding to the other user device to the user device within a second predetermined time range, and then the user device decides to independently participate in the federated learning.

24. The electronic device according to any one of claims 1 to 23, wherein the processing circuit is configured to transmit information regarding the federative learning to the user device and the linked user device when the user device decides to participate in the federative learning independently and selects a linked user device to be trained in cooperation with the user device.

25. The electronic device according to claim 24, wherein the processing circuit is configured to broadcast the training target model and the division points for dividing the training target model to obtain the divided models to the user device and the linked user device.

26. The electronic device according to claim 24, wherein the processing circuit is configured to transmit a segmented model corresponding to the user device to the user device and transmit a segmented model corresponding to the linked user device to the linked user device.

27. The electronic device according to any one of claims 24 to 26, wherein the processing circuit is configured such that the user device and the linked user device each report a segmented model after training, and the electronic device notifies the user device to combine the segmented models after training to obtain a complete model after training.

28. The electronic device according to any one of claims 24 to 26, wherein the processing circuit is configured to notify the user device or the linked user device to combine the trained segmented models to obtain a complete trained model and report the complete trained model to the electronic device.

29. The electronic device according to any one of claims 24 to 28, wherein the processing circuit is configured to notify the user device and the linked user device of information regarding a side link between the user device and the linked user device.

30. Electronic equipment for wireless communication, An electronic device including a processing circuit configured to report the self-state information of the electronic device to the network-side device, which provides services to the electronic device, so that the network-side device determines whether the electronic device will participate in the federative learning in a non-independent manner by training the divided models obtained by dividing the model to be trained in cooperation with other electronic devices, or whether it will participate in the federative learning independently by training the model to be trained independently.

31. The electronic device according to claim 30, wherein the self-state information includes first channel state information of the link between the electronic device and the network-side device.

32. The electronic device according to claim 31, wherein the first channel state information includes at least one of the signal-to-interference noise ratio SINR of the uplink between the electronic device and the network-side device, reference signal received power RSRP, reference signal received quality RSRQ, and reference signal received intensity RSRI.

33. The electronic device according to claim 30, wherein the self-state information includes computing and storage capability information of the electronic device.

34. The electronic device according to claim 33, wherein the computing and storage capability information includes at least one of the CPU usage rate and memory size of the electronic device.

35. The electronic device according to claim 30, wherein the self-state information includes local data information of the electronic device.

36. The electronic device according to claim 35, wherein the local data information includes the number of training samples used when the electronic device independently participates in the federative learning, and the sample dimensions.

37. The electronic device according to claim 30, wherein the self-state information includes power consumption information of the electronic device.

38. The electronic device according to claim 30, wherein the self-state information includes the location information of the electronic device.

39. The electronic device according to claim 30, wherein the self-state information includes the movement information of the electronic device.

40. The electronic device according to claim 39, wherein the movement information includes at least one of the movement speed, movement direction, and stay time of the electronic device.

41. The electronic device according to claim 30, wherein the self-state information includes second channel state information of a side link between the electronic device and other electronic devices within a predetermined range of the electronic device.

42. The aforementioned processing circuit is When participating in the aforementioned federative learning independently, in each round of the global training of the aforementioned federative learning, with respect to the first segmented model corresponding to the electronic device, The first data obtained by training the first segmentation model is transmitted to a linked electronic device that trains in conjunction with the electronic device. The linked electronic device updates the first partition model based on the second data received from the linked electronic device, which is obtained by training a second partition model corresponding to the linked electronic device based on the first data. The electronic device according to any one of claims 30 to 41, configured to obtain a first split model after training by performing the following a predetermined number of times.

43. The processing circuit is configured to report the first segmented model after training to the network-side device so that the network-side device can combine the first segmented model after training with the second segmented model after training received from the linked electronic device to obtain a complete model after training. The electronic device according to claim 42, wherein the second segmented model after training is obtained by the collaborative electronic device training the second segmented model in the global training.

44. The processing circuit is configured to combine the first partitioned model after training and the second partitioned model after training to obtain a complete model after training, and to report the complete model after training to the network-side device. The electronic device according to claim 42, wherein the second segmented model after training is obtained by the collaborative electronic device training the second segmented model in the global training.

45. Electronic equipment for wireless communication, An electronic device including a processing circuit configured to report self-state information of the electronic device to the network-side device providing services to the electronic device, so as to determine whether the electronic device can train the divided models obtained by dividing the model to be trained in cooperation with other electronic devices during federated learning, and whether the other electronic devices participate in the federated learning independently or whether the other electronic devices train the model to be trained independently and participate in the federated learning independently.

46. The electronic device according to claim 45, wherein the self-state information includes third channel state information of the link between the electronic device and the network-side device.

47. The electronic device according to claim 46, wherein the third channel state information includes at least one of the signal-to-interference noise ratio SINR of the uplink between the electronic device and the network-side device, reference signal received power RSRP, reference signal received quality RSRQ, and reference signal received intensity RSRI.

48. The electronic device according to claim 45, wherein the self-state information includes computing and storage capability information of the electronic device.

49. The electronic device according to claim 48, wherein the computing and storage capability information includes at least one of the CPU usage rate and memory size of the electronic device.

50. The electronic device according to claim 49, wherein the self-state information includes local data information of the electronic device.

51. The electronic device according to claim 50, wherein the local data information includes the number of training samples used when the electronic device independently participates in the federative learning, and the sample dimensions.

52. The electronic device according to claim 45, wherein the self-state information includes power consumption information of the electronic device.

53. The electronic device according to claim 45, wherein the self-state information includes the location information of the electronic device.

54. The electronic device according to claim 45, wherein the self-state information includes the movement information of the electronic device.

55. The electronic device according to claim 54, wherein the movement information includes at least one of the movement speed, movement direction, and stay time of the electronic device.

56. The electronic device according to claim 45, wherein the self-state information includes a fourth channel state information of a side link between the electronic device and other electronic devices within a predetermined range of the electronic device.

57. The aforementioned processing circuit is If the other electronic devices participate in the federative learning independently, in each round of the global training of the federative learning, with respect to the second partition model corresponding to the electronic device, The system receives first data obtained by training a first segmentation model corresponding to the other electronic device from the other electronic device. The second partition model is trained based on the first data to obtain second data, and the second data is transmitted to the other electronic device so that the other electronic device updates the first partition model. The electronic device according to any one of claims 45 to 56, configured to obtain a second split model after training by performing the following a predetermined number of times.

58. The processing circuit is configured to report the trained second segmented model to the network-side device so that the network-side device can combine the trained second segmented model with the trained first segmented model received from the other electronic device to obtain a complete trained model. The electronic device according to claim 57, wherein the first segmented model after training is obtained by the other electronic device training the first segmented model in the global training.

59. The processing circuit is configured to combine the second partitioned model after training and the first partitioned model after training to obtain a complete model after training, and to report the complete model after training to the network-side device. The electronic device according to claim 57, wherein the first segmented model after training is obtained by the other electronic device training the first segmented model in the global training.

60. The electronic device according to any one of claims 1 to 29, The electronic device according to any one of claims 30 to 44, A wireless communication system comprising an electronic device according to any one of claims 45 to 59.

61. A method for wireless communication, A method comprising determining, in federated learning, whether a user device participates in the federated learning non-independently by training a divided model obtained by dividing the model to be trained in cooperation with other user devices, or independently by training the model to be trained independently, based on self-state information reported by a user device within the service scope of an electronic device.

62. A method for wireless communication, A method comprising reporting the self-state information of an electronic device to the network-side device providing services to an electronic device, so that the electronic device can determine whether to participate in the federative learning in a non-independent manner by training the divided models obtained by dividing the model to be trained in cooperation with other electronic devices, or to participate in the federative learning independently by training the model to be trained independently.

63. A method for wireless communication, A method comprising a network-side device providing services to an electronic device reporting the self-state information of the electronic device to the network-side device to determine whether the electronic device can train the partitioned models obtained by partitioning the model to be trained in cooperation with other electronic devices in federated learning, and whether the other electronic devices participate in the federated learning independently or whether the other electronic devices train the model to be trained independently and participate in the federated learning independently.

64. A computer-readable storage medium having stored computer-executable instructions that, if executed, cause a method for wireless communication according to any one of claims 61 to 63.

Images