Federated learning management method and apparatus, device, readable storage medium, and program

Abstract

The present disclosure provides a federated learning management method and apparatus, a device, a readable storage medium, and a program. The method applied to a wireless intelligent controller comprises: acquiring computing power information and / or attribute information of a first device, wherein the first device includes a base station and / or a terminal; and on the basis of the computing power information and / or the attribute information, sending federated learning task related information to the first device, wherein the federated learning task related information comprises a federated learning task, or the federated learning task related information is used for determining the federated learning task.

Description

Federated learning management method, device, equipment, readable storage medium and program

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to Chinese Patent Application No. 202410183385.3 filed in China on February 19, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present disclosure relates to the field of wireless access network communication technology, and in particular to a federated learning management method, apparatus, device, readable storage medium, and program. Background Art

[0004] In wireless communication networks, as services become increasingly diverse and demands are increasingly met, intelligence is gradually becoming prevalent in the network. Open and intelligent radio access networks (O-RAN) are increasingly widely used. However, related technologies are still unable to efficiently and reasonably utilize the computing power resources of base stations and / or terminals through the O-RAN architecture for intelligent training and inference, while also meeting the real-time requirements of artificial intelligence (AI) applications. Summary of the Invention

[0005] The purpose of the present disclosure is to provide a federated learning management method, apparatus, device, readable storage medium, and program to address the problem of low computing resource utilization efficiency and inability to intelligently manage base stations and terminals under the O-RAN architecture.

[0006] In order to solve the above technical problems, the present disclosure provides a federated learning management method applied to a wireless intelligent controller, including:

[0007] Obtain computing power information and / or attribute information of a first device, where the first device includes a base station and / or a terminal;

[0008] According to the computing power information and / or attribute information, federated learning task related information is sent to the first device, where the federated learning task related information includes a federated learning task, or the federated learning task related information is used to determine the federated learning task.

[0009] Optionally, the attribute information of the terminal includes one or more of the following:

[0010] Terminal location information, terminal type, service type, and service priority.

[0011] Optionally, the sending, to the first device, information related to the federated learning task according to the computing power information and / or attribute information includes:

[0012] Determining, based on the computing power information and / or the attribute information, information related to a federated learning task corresponding to the terminal, where the information related to the federated learning task corresponding to the terminal includes a federated learning task allocation strategy, where the federated learning task allocation strategy is used to determine the federated learning task;

[0013] Sending information related to the federated learning task corresponding to the terminal to the base station.

[0014] Optionally, after sending the federated learning task related information corresponding to the terminal to the base station, the method further includes:

[0015] Receive a federated learning task allocation result sent by the base station, where the federated learning task allocation result is determined according to the federated learning task allocation strategy, and the federated learning task allocation result includes: a terminal that executes a federated learning task and a federated learning task corresponding to each terminal.

[0016] Optionally, the sending, to the first device, information related to the federated learning task according to the computing power information and / or attribute information includes:

[0017] Determine, based on the computing power information and / or attribute information of the terminal, information related to the federated learning task corresponding to the terminal, where the information related to the federated learning task corresponding to the terminal includes: the terminal executing the federated learning task and the federated learning task corresponding to each terminal;

[0018] The federated learning task related information corresponding to the terminal is sent to the terminal through the base station.

[0019] Optionally, the sending, to the first device, information related to the federated learning task according to the computing power information and / or attribute information includes:

[0020] Determine, based on the computing power information and / or attribute information of the base station, information related to the federated learning task corresponding to the base station, where the information related to the federated learning task corresponding to the base station includes: the base station that performs the federated learning task and the federated learning task corresponding to each base station;

[0021] Sending information related to the federated learning task corresponding to the base station to the base station.

[0022] Optionally, after sending the federated learning task related information to the first device, the method further includes:

[0023] Obtaining learning indicator information sent by the first device, where the learning indicator information is obtained by the first device performing the federated learning task;

[0024] According to the learning indicator information, an adjustment strategy corresponding to the federated learning task is sent to the first device, or a model is assigned to the inference model obtained by the first device executing the federated learning task.

[0025] Optionally, before obtaining the learning indicator information sent by the first device, the method further includes:

[0026] Sending a subscription request message to the base station, where the subscription request message is used to request obtaining the learning indicator information;

[0027] A subscription response message is received from the base station according to the subscription request message, where the subscription response message is used to confirm reporting of the learning indicator information.

[0028] Optionally, the subscription request message carries reporting conditions for learning indicator information;

[0029] Obtaining learning indicator information sent by the first device, including:

[0030] Obtain the learning indicator information sent by the base station according to the reporting condition.

[0031] Optionally, sending the adjustment strategy corresponding to the federated learning task to the first device according to the learning indicator information includes:

[0032] When it is determined that the learning indicator information satisfies the first preset condition, determining an adjustment strategy corresponding to the federated learning task;

[0033] Sending an adjustment strategy corresponding to the federated learning task to the first device.

[0034] Optionally, the adjustment strategy corresponding to the federated learning task includes one or more of the following:

[0035] Increase federated learning tasks, reduce federated learning tasks, increase the first device that executes federated learning tasks, and reduce the first device that executes federated learning tasks.

[0036] Optionally, the learning indicator information includes one or more of the following:

[0037] Model training speed, computing power consumption, and model training effect;

[0038] The first preset condition includes:

[0039] The model training speed is less than a first threshold and / or the consumed computing power is greater than a second threshold and / or the model training effect does not meet the preset effect.

[0040] Optionally, the performing model allocation on the inference model obtained by the first device performing the federated learning task according to the learning indicator information includes:

[0041] If it is determined that the learning indicator information satisfies a second preset condition, determining the currently obtained aggregate model as an inference model, where the aggregate model is obtained by aggregating models generated by the first device when executing the federated learning task;

[0042] A model assignment is performed on the inference model.

[0043] Optionally, performing model allocation on the inference model includes:

[0044] Obtaining task requirement information of the business to be processed, wherein the task requirement information includes real-time requirements and / or input and output requirements;

[0045] The reasoning model is assigned a model according to the task requirement information.

[0046] Optionally, after sending the federated learning task related information to the first device, the method further includes:

[0047] Obtaining first eigenvalue information sent by the first device, where the first eigenvalue information includes sub-eigenvalues ​​of a model when the first device performs the federated learning task;

[0048] According to the first eigenvalue information, first indication information is sent to the first device, where the first indication information is used to instruct the first device to update the sub-eigenvalue of the model.

[0049] The present disclosure also provides a federated learning management method, which is applied to a base station and includes:

[0050] Receiving federated learning task related information sent by the wireless intelligent controller, wherein the federated learning task related information includes a federated learning task or a federated learning task allocation strategy;

[0051] In a case where the federated learning task-related information includes the federated learning task, executing the federated learning task;

[0052] In a case where the federated learning task related information includes the federated learning task allocation strategy, determining a federated learning task allocation result corresponding to the terminal according to the federated learning task allocation strategy, the federated learning task allocation result including the terminal that performs the federated learning task and the federated learning task corresponding to each terminal;

[0053] The federated learning task allocation result is sent to the wireless intelligent controller, and the corresponding federated learning task is sent to the terminal according to the federated learning task allocation result.

[0054] Optionally, after executing the federated learning task, the method further includes:

[0055] Sending learning indicator information to the wireless intelligent controller, where the learning indicator information is obtained by executing the federated learning task.

[0056] Optionally, after sending the corresponding federated learning task to the terminal according to the federated learning task allocation result, the method further includes:

[0057] Obtaining learning indicator information corresponding to the execution of the federated learning task by the terminal;

[0058] The learning indicator information is sent to the wireless intelligent controller.

[0059] Optionally, before sending the learning indicator information to the wireless intelligent controller, the method further includes:

[0060] receiving a subscription request message sent by the wireless intelligent controller, where the subscription request message is used to request acquisition of the learning indicator information;

[0061] A subscription response message is sent to the wireless intelligent controller according to the subscription request message, where the subscription response message is used to confirm reporting of the learning indicator information.

[0062] Optionally, the subscription request message carries a reporting condition for learning indicator information;

[0063] The learning indicator information is sent to the wireless intelligent controller according to the reporting condition.

[0064] Optionally, after sending the learning indicator information to the wireless intelligent controller, the method further includes:

[0065] Receiving an adjustment strategy corresponding to the federated learning task sent by the wireless intelligent controller;

[0066] Adjustments are made according to the adjustment strategy corresponding to the federated learning task.

[0067] Optionally, the method further includes:

[0068] Sending first eigenvalue information to the wireless intelligent controller, where the first eigenvalue information includes a sub-eigenvalue of a model when the base station performs the federated learning task;

[0069] receiving first indication information fed back by the wireless intelligent controller according to the first characteristic value information;

[0070] The sub-feature values ​​of the model are updated according to the first indication information.

[0071] Optionally, the method further includes:

[0072] Obtaining a sub-eigenvalue of a model sent by the terminal when the terminal executes the corresponding federated learning task;

[0073] Sending first eigenvalue information to the wireless intelligent controller, where the first eigenvalue information includes a sub-eigenvalue of a model when the terminal performs the federated learning task;

[0074] receiving first indication information fed back by the wireless intelligent controller according to the first characteristic value information;

[0075] Update information is sent to the terminal according to the first indication information, where the update information is used to instruct the terminal to update the sub-feature value of the model.

[0076] The present disclosure also provides a federated learning management method, which is applied to a terminal and includes:

[0077] Receiving federated learning task related information sent by a wireless intelligent controller through a base station, wherein the federated learning task related information includes a federated learning task, or receiving a federated learning task sent by the base station;

[0078] Execute the federated learning task.

[0079] Optionally, after executing the federated learning task, the method further includes:

[0080] Sending learning indicator information to the wireless intelligent controller through a base station or sending the learning indicator information to the base station;

[0081] The learning indicator information is obtained by executing the federated learning task.

[0082] Optionally, the method further includes:

[0083] Sending the sub-eigenvalues ​​of the model when executing the federated learning task to the base station;

[0084] Obtaining update information sent by the base station;

[0085] The sub-feature values ​​of the model are updated according to the update information.

[0086] An embodiment of the present disclosure further provides a network device, including a transceiver and a processor, wherein the transceiver is configured to:

[0087] Obtain computing power information and / or attribute information of a first device, where the first device includes a base station and / or a terminal;

[0088] According to the computing power information and / or attribute information, federated learning task related information is sent to the first device, where the federated learning task related information includes a federated learning task, or the federated learning task related information is used to determine the federated learning task.

[0089] The present disclosure also provides a base station, including a transceiver and a processor, wherein:

[0090] The transceiver is used to: receive federated learning task related information sent by the wireless intelligent controller, wherein the federated learning task related information includes a federated learning task or a federated learning task allocation strategy;

[0091] The processor is configured to: execute the federated learning task when the federated learning task related information includes the federated learning task;

[0092] The processor is configured to: when the federated learning task related information includes the federated learning task allocation strategy, determine a federated learning task allocation result corresponding to the terminal according to the federated learning task allocation strategy, wherein the federated learning task allocation result includes a terminal that performs the federated learning task and a federated learning task corresponding to each terminal;

[0093] The transceiver is used to: send the federated learning task allocation result to the wireless intelligent controller, and send the corresponding federated learning task to the terminal according to the federated learning task allocation result.

[0094] An embodiment of the present disclosure further provides a terminal, including a transceiver and a processor, wherein:

[0095] The transceiver is configured to: receive federated learning task related information sent by a wireless intelligent controller through a base station, wherein the federated learning task related information includes a federated learning task, or receive a federated learning task sent by the base station;

[0096] The processor is used to: execute the federated learning task.

[0097] The present disclosure also provides a federated learning management method and apparatus, which is applied to an intelligent controller and includes:

[0098] A first acquisition module is configured to acquire computing power information and / or attribute information of a first device, where the first device includes a base station and / or a terminal;

[0099] A first sending module is used to send federated learning task related information to the first device based on the computing power information and / or attribute information, where the federated learning task related information includes the federated learning task, or the federated learning task related information is used to determine the federated learning task.

[0100] The present disclosure also provides a federated learning management method and apparatus, which is applied to a base station and includes:

[0101] A first receiving module is configured to receive federated learning task related information sent by the wireless intelligent controller, wherein the federated learning task related information includes a federated learning task or a federated learning task allocation strategy;

[0102] a first execution module, configured to execute the federated learning task if the federated learning task related information includes the federated learning task;

[0103] a first determining module configured to, when the federated learning task related information includes the federated learning task allocation strategy, determine a federated learning task allocation result corresponding to the terminal according to the federated learning task allocation strategy, the federated learning task allocation result including the terminal that performs the federated learning task and the federated learning task corresponding to each terminal;

[0104] The second sending module is used to send the federated learning task allocation result to the wireless intelligent controller, and send the corresponding federated learning task to the terminal according to the federated learning task allocation result.

[0105] The present disclosure also provides a federated learning management method and apparatus, which is applied to a terminal and includes:

[0106] A second receiving module is configured to receive federated learning task related information sent by the wireless intelligent controller through the base station, wherein the federated learning task related information includes the federated learning task, or receive the federated learning task sent by the base station;

[0107] The second execution module is used to execute the federated learning task.

[0108] An embodiment of the present disclosure further provides a network device, comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the federated learning management method as described in any one of the above items.

[0109] An embodiment of the present disclosure further provides a readable storage medium, including: a program stored on the readable storage medium, and when the program is executed by a processor, the steps of the federated learning management method as described in any one of the above items are implemented.

[0110] An embodiment of the present disclosure further provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the steps of the federated learning management method as described in any one of the above items.

[0111] At least one of the above technical solutions of the present disclosure has the following beneficial effects:

[0112] In the above scheme, the wireless intelligent controller sends information related to the federated learning task to the first device based on the computing power information and / or attribute information of the first device, wherein the first device includes a base station and / or a terminal, and the base station and / or the terminal performs the corresponding federated learning task based on the information related to the federated learning task; the wireless intelligent controller intelligently manages the federated learning tasks and allocates the federated learning tasks to be performed by the base station and / or the terminal, thereby realizing the application of federated learning under the O-RAN architecture and improving the computing power resource utilization efficiency of the base station and the terminal. BRIEF DESCRIPTION OF THE DRAWINGS

[0113] FIG1 is a flow chart of a federated learning management method according to one embodiment of the present disclosure;

[0114] FIG2 is a flow chart of a federated learning management method according to another embodiment of the present disclosure;

[0115] FIG3 is a flow chart of a federated learning management method according to another embodiment of the present disclosure;

[0116] FIG4 is a schematic diagram of a process flow of one embodiment of the method according to the present disclosure;

[0117] FIG5 is a flow chart of another embodiment of the method according to the embodiment of the present disclosure;

[0118] FIG6 is a flow chart of another embodiment of the method according to the embodiment of the present disclosure;

[0119] FIG7 is a schematic diagram of the structure of a network device according to an embodiment of the present disclosure;

[0120] FIG8 is a schematic structural diagram of a base station according to an embodiment of the present disclosure;

[0121] FIG9 is a schematic structural diagram of a terminal according to an embodiment of the present disclosure;

[0122] FIG10 is a schematic diagram of the structure of a federated learning management device according to one embodiment of the present disclosure;

[0123] FIG11 is a schematic structural diagram of a federated learning management device according to another embodiment of the present disclosure;

[0124] FIG12 is a schematic structural diagram of a federated learning management device according to another embodiment of the present disclosure. DETAILED DESCRIPTION

[0125] The following will be combined with the accompanying drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the embodiments described are only part of the embodiments of the present disclosure and not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0126] Compared to traditional radio access network architecture, the Open-Radio Access Network (O-RAN) proposes a wireless intelligent controller, which includes non-real-time wireless intelligent controllers and near-real-time wireless intelligent controllers.

[0127] Among them, the Near Real-Time RAN Intelligence Controller (Near-RT RIC) is divided into platform functions and x application (Application, App) instance application functions. The platform function is used to support the deployment and operation of instance applications, and is divided into modules such as conflict resolution, subscription management, management services, artificial intelligence (Artificial Intelligence, AL) / machine learning (Machine Learning, ML) support, and security; among them, "AI / ML support" is an important function in the O-RAN intelligent architecture, which is used to support the AI ​​function of intelligent applications and realize real-time intelligent control of base stations / terminals. The rApp application for deploying and managing federated learning in the non-real-time intelligent controller can deploy different rApp applications to manage different AI applications according to business characteristics, AI applications, AI attributes, etc. The present disclosure introduces the functions and processes for managing federated learning applications and AI reasoning in the O-RAN architecture.

[0128] As shown in FIG1 , an embodiment of the present disclosure provides a federated learning management method applied to a wireless intelligent controller, including:

[0129] Step S101: Obtain computing power information and / or attribute information of a first device, where the first device includes a base station and / or a terminal;

[0130] In step S101, the wireless intelligent controller can be obtained by subscription through the management service module, or by subscription through the base station or terminal, or partly by the management service module and partly by subscription through the base station or terminal.

[0131] Step S102: Send federated learning task related information to the first device based on the computing power information and / or attribute information, where the federated learning task related information includes the federated learning task, or the federated learning task related information is used to determine the federated learning task.

[0132] In step S102, the federated learning task requires multiple base stations and / or terminals to complete federated training. The wireless intelligent controller allocates the federated learning task based on the computing power information and / or attribute information of the first device, so that each base station and / or terminal corresponds to a subtask, namely the above-mentioned federated learning task.

[0133] In the disclosed embodiment, the wireless intelligent controller sends information related to the federated learning task to the first device based on the computing power information and / or attribute information of the first device, allocates the federated learning tasks, and realizes intelligent management of the federated learning tasks by the wireless intelligent controller, thereby realizing the application of federated learning under the O-RAN architecture and improving the computing power resource utilization efficiency of base stations and terminals.

[0134] Optionally, the attribute information of the terminal includes one or more of the following:

[0135] Terminal location information, terminal type, service type, and service priority.

[0136] In the disclosed embodiment, the attribute information of the terminal includes, but is not limited to, one or more of the terminal location information, terminal type, service type, and service priority. When the wireless intelligent controller assigns a federal task to the terminal, it more reasonably utilizes the terminal's attributes, and the terminal performs the task more efficiently.

[0137] Optionally, the sending, to the first device, information related to the federated learning task according to the computing power information and / or attribute information includes:

[0138] Determining, based on the computing power information and / or the attribute information, information related to a federated learning task corresponding to the terminal, where the information related to the federated learning task corresponding to the terminal includes a federated learning task allocation strategy, where the federated learning task allocation strategy is used to determine the federated learning task;

[0139] Sending information related to the federated learning task corresponding to the terminal to the base station.

[0140] In an embodiment of the present disclosure, one implementation method of sending federated learning task-related information by the wireless intelligent controller to perform federated learning task allocation in step S102 is described. When the wireless intelligent controller obtains the computing power information and / or attribute information of the terminal, it determines the federated learning task allocation strategy corresponding to the terminal based on the above computing power information and / or attribute information. The federated learning task allocation strategy may include: the number of terminal selections, the terminal requirements corresponding to the federated learning task, such as the computing power size of the terminal, etc.

[0141] The wireless intelligent controller sends federated learning task-related information to the base station. The above-mentioned related information includes the terminal corresponding federated learning task allocation strategy and the total federated learning task, which is used to instruct the base station to allocate terminals and federated learning tasks.

[0142] Optionally, after sending the federated learning task related information corresponding to the terminal to the base station, the method further includes:

[0143] Receive a federated learning task allocation result sent by the base station, where the federated learning task allocation result is determined according to the federated learning task allocation strategy, and the federated learning task allocation result includes: a terminal that executes a federated learning task and a federated learning task corresponding to each terminal.

[0144] In the embodiment of the present disclosure, after the base station assigns the federated learning task to the terminal according to the federated learning task assignment strategy, it feeds back the federated learning task assignment result to the wireless intelligent controller. The federated learning task assignment result can be represented in a one-to-one correspondence between the terminal identifier executing the federated learning task and the federated learning task.

[0145] The wireless intelligent controller instructs the base station to perform task allocation, and finally obtains the federated learning task allocation results to perform intelligent management of the federated learning tasks.

[0146] Optionally, the sending, to the first device, information related to the federated learning task according to the computing power information and / or attribute information includes:

[0147] Determine, based on the computing power information and / or attribute information of the terminal, information related to the federated learning task corresponding to the terminal, where the information related to the federated learning task corresponding to the terminal includes: the terminal executing the federated learning task and the federated learning task corresponding to each terminal;

[0148] The federated learning task related information corresponding to the terminal is sent to the terminal through the base station.

[0149] In an embodiment of the present disclosure, one implementation method of sending federated learning task-related information by the wireless intelligent controller to perform federated learning task allocation in step S102 is described. When the wireless intelligent controller obtains the computing power information and / or attribute information of the terminal, the wireless intelligent controller allocates federated learning tasks based on the above computing power information and / or attribute information, determines the terminal that executes the federated learning task and the federated learning task corresponding to each terminal, and performs task allocation and intelligent management.

[0150] Optionally, the sending, to the first device, information related to the federated learning task according to the computing power information and / or attribute information includes:

[0151] Determine, based on the computing power information and / or attribute information of the base station, information related to the federated learning task corresponding to the base station, where the information related to the federated learning task corresponding to the base station includes: the base station that performs the federated learning task and the federated learning task corresponding to each base station;

[0152] Sending information related to the federated learning task corresponding to the base station to the base station.

[0153] In an embodiment of the present disclosure, one implementation method of sending federated learning task-related information by the wireless intelligent controller to allocate federated learning tasks in step S102 is described. When the wireless intelligent controller obtains the computing power information and / or attribute information of the base station, the wireless intelligent controller allocates federated learning tasks according to the above computing power information and / or attribute information, determines the base station that executes the federated learning task and the federated learning task corresponding to each base station, and performs task allocation and intelligent management.

[0154] Optionally, after sending the federated learning task related information to the first device, the method further includes:

[0155] Obtaining learning indicator information sent by the first device, where the learning indicator information is obtained by the first device performing the federated learning task;

[0156] According to the learning indicator information, an adjustment strategy corresponding to the federated learning task is sent to the first device, or a model is assigned to the inference model obtained by the first device executing the federated learning task.

[0157] In an embodiment of the present disclosure, when the base station and / or terminal performs the corresponding federated learning task, the wireless intelligent controller monitors the base station and / or the terminal, and obtains learning indicator information of the base station and / or the terminal performing the corresponding federated learning task, wherein the learning indicator information includes the federated service quality (Quality of Service, QoS) and / or key performance indicator (Key Performance Indicator, KPI) of the first device performing federated learning, specifically, including information such as the power (energy) consumed by training and computing power, computing time, federated learning model size, model accuracy, training speed, feedback cycle, etc.; the federated learning task is adjusted according to the learning indicator information, and the learning model distribution, federated task allocation, and terminals and base stations participating in the training in related technologies are optimized, or the learning indicator information is sent to the inference model to optimize the terminals and base stations participating in the inference.

[0158] Optionally, before obtaining the learning indicator information sent by the first device, the method further includes:

[0159] Sending a subscription request message to the base station, where the subscription request message is used to request obtaining the learning indicator information;

[0160] A subscription response message is received from the base station according to the subscription request message, where the subscription response message is used to confirm reporting of the learning indicator information.

[0161] In the embodiment of the present disclosure, the learning indicator information of the first device is obtained through subscription. The wireless intelligent controller subscribes to the learning indicator information from the base station, and the base station feeds back a subscription response message after agreeing to the subscription request. Generally, the base station collects the learning indicator information of the terminal in real time, and the subscription request message is used to request the base station to report the learning indicator information of the base station and the terminal.

[0162] In addition, the wireless intelligent controller can subscribe to the learning indicator information of a task based on an AI application / computing task as an index.

[0163] Optionally, the subscription request message carries reporting conditions for learning indicator information;

[0164] Obtaining learning indicator information sent by the first device, including:

[0165] Receive the learning indicator information sent by the base station according to the reporting condition.

[0166] In an embodiment of the present disclosure, the base station collects learning indicator information of the terminal in real time, and the reporting conditions include periodic reporting or quantity reporting. When the reporting condition is periodic reporting, the base station reports all the collected learning indicator information to the wireless intelligent controller in each period. At the same time, for the information of the terminal that is not reported, the base station can indicate the reason for non-reporting, including but not limited to inability to obtain, incomplete training task, etc.; when the reporting condition is quantity reporting, after collecting the learning indicator information of the terminals that meet the quantity, the base station reports the collected learning indicator information to the wireless intelligent controller, where the number of terminals in the quantity report can be an absolute number or a certain percentage of terminals.

[0167] Optionally, sending the adjustment strategy corresponding to the federated learning task to the first device according to the learning indicator information includes:

[0168] When it is determined that the learning indicator information satisfies the first preset condition, determining an adjustment strategy corresponding to the federated learning task;

[0169] Sending an adjustment strategy corresponding to the federated learning task to the first device.

[0170] In the embodiment of the present disclosure, the learning indicator information meets the first preset condition, indicating that the federated learning task needs to be optimized. Therefore, the adjustment strategy corresponding to the federated learning task is determined and sent to the first device side for adjustment, thereby realizing intelligent management of the federated learning task by the wireless intelligent controller.

[0171] The wireless intelligent controller can directly send the adjustment strategy corresponding to the federated learning task to the base station to adjust the federated learning task executed by the base station; the wireless intelligent controller can also directly send the adjustment strategy corresponding to the federated learning task to the terminal to adjust the federated learning task executed by the terminal; the wireless intelligent controller can also send the adjustment strategy corresponding to the federated learning task to the base station to enable the base station to adjust the federated learning task executed by the terminal.

[0172] Optionally, when the attribute information and / or computing power information of the first device is monitored to be updated, determining an adjustment strategy corresponding to the federated learning task;

[0173] Sending an adjustment strategy corresponding to the federated learning task to the first device.

[0174] In the embodiment of the present disclosure, if the attribute information and / or computing power information of the first device is updated, the initial federated learning task will no longer be applicable and the federated learning task needs to be adjusted. Therefore, the adjustment strategy corresponding to the federated learning task is determined and sent to the first device side for adjustment, thereby realizing intelligent management of the federated learning task by the wireless intelligent controller.

[0175] Optionally, the learning indicator information includes one or more of the following:

[0176] Model training speed, computing power consumption, and model training effect;

[0177] The first preset condition includes:

[0178] The model training speed is less than a first threshold and / or the consumed computing power is greater than a second threshold and / or the model training effect does not meet the preset effect.

[0179] In the disclosed embodiment, if the model training speed is too low or the computing power consumption is too high, it indicates that the federated learning task is not well matched with the corresponding base station or terminal, or that the task volume of the federated learning task is too large, and the federated learning task needs to be adjusted; the model training effect includes model accuracy, and the model accuracy includes but is not limited to one or more of the model accuracy, precision and recall rate. If the model training effect does not meet the preset effect, it means that the model training needs to be continued, the federated learning task needs to be adjusted, and a new round of tasks needs to be initiated.

[0180] Optionally, the adjustment strategy corresponding to the federated learning task includes one or more of the following:

[0181] Increase federated learning tasks, reduce federated learning tasks, increase the first device that executes federated learning tasks, and reduce the first device that executes federated learning tasks.

[0182] In the disclosed embodiment, the adjustment strategy corresponding to the federated learning task is explained. If the model training speed is too low or the computing power consumption is too high, the federated learning task needs to be adjusted. For example, the first device that executes the federated learning task can be increased or the federated tasks of the first device can be reduced to allow other first devices with idle computing power to execute. If the model training effect is not satisfactory to the preset effect, the federated learning task needs to be adjusted and a new round of tasks needs to be initiated.

[0183] Optionally, the performing model allocation on the inference model obtained by the first device performing the federated learning task according to the learning indicator information includes:

[0184] If it is determined that the learning indicator information satisfies a second preset condition, determining the currently obtained aggregate model as an inference model, where the aggregate model is obtained by aggregating models generated by the first device when executing the federated learning task;

[0185] A model assignment is performed on the inference model.

[0186] In the embodiment of the present disclosure, after the federated learning task training is completed or the number of iterations is reached, it is determined whether the KPI in the learning indicator information meets the second preset condition. If so, the aggregated model currently obtained through model training is determined as the inference model;

[0187] Model distribution can directly send the inference model to the wireless intelligent controller side, base station side, or terminal side. It can also split the inference model, with sub-model 1 distributed to the wireless intelligent controller side, 2 distributed to the base station side, and 3 distributed to the terminal side. Alternatively, the base station can obtain the inference model / sub-model and further split the model and distribute it to one or more terminals.

[0188] Optionally, performing model assignment on the inference model includes:

[0189] Obtaining task requirement information of the business to be processed, wherein the task requirement information includes real-time requirements and / or input and output requirements;

[0190] The reasoning model is assigned a model according to the task requirement information.

[0191] In the embodiment of the present disclosure, the task requirement information includes but is not limited to one or more of real-time requirements and input and output requirements. According to the task requirement information of the business to be processed, it is determined whether the inference task corresponding to the business to be processed is executed on the base station, terminal or wireless intelligent controller side. According to the execution end of the inference task, the inference model is assigned. In the case where the task requirement information includes real-time requirements, if the real-time requirements are higher, the inference task is assigned to the base station side or the terminal side; if the real-time requirements are lower, the inference task is assigned to the wireless intelligent controller side.

[0192] Optionally, after sending the federated learning task related information to the first device, the method further includes:

[0193] Obtaining first eigenvalue information sent by the first device, where the first eigenvalue information includes sub-eigenvalues ​​of a model when the first device performs the federated learning task;

[0194] According to the first eigenvalue information, first indication information is sent to the first device, where the first indication information is used to instruct the first device to update the sub-eigenvalue of the model.

[0195] In the embodiment of the present disclosure, the wireless intelligent controller can also intelligently jointly update the training model. During the process of the first device executing the federated learning task to perform model training, the sub-feature value of the model of the first device is obtained, and the wireless intelligent controller determines whether the sub-feature value of the first device needs to be updated. Generally, the wireless intelligent controller needs to refer to the learning indicator information to make a judgment. If an update is required, the wireless intelligent controller sends a first indication information to the first device, instructing the first device to update the sub-feature value, wherein the first indication information can be in the form of a list, including the device to be updated and the corresponding updated sub-feature value, wherein the terminal corresponding to the same sub-feature value can be identified by the feature value identifier.

[0196] It should be noted that during the federated training process, updating the sub-feature values ​​of the model means updating the corresponding training model.

[0197] Optionally, after sending the first indication information to the first device, the method further includes:

[0198] Obtain the aggregation model sent by the base station;

[0199] The federated learning task related information is updated according to the aggregation model.

[0200] In the embodiment of the present disclosure, the wireless intelligent controller receives the aggregation model sent by the base station, updates the local aggregation model, and updates the federated learning task-related information according to the updated aggregation model, such as the federated learning content or the federated learning allocation strategy of the terminal. The management and allocation method of the updated federated learning task-related information is the same as the management and allocation method of the previous federated learning task-related information, and will not be repeated here.

[0201] As shown in FIG2 , the embodiment of the present disclosure further provides a federated learning management method, which is applied to a base station and includes:

[0202] Step S201: receiving federated learning task related information sent by a wireless intelligent controller, wherein the federated learning task related information includes a federated learning task or a federated learning task allocation strategy;

[0203] Step S202: executing the federated learning task if the federated learning task-related information includes the federated learning task;

[0204] Step S203: When the federated learning task-related information includes the federated learning task allocation strategy, determining a federated learning task allocation result corresponding to the terminal according to the federated learning task allocation strategy, wherein the federated learning task allocation result includes the terminal that performs the federated learning task and the federated learning task corresponding to each terminal;

[0205] Step S204: Send the federated learning task allocation result to the wireless intelligent controller, and send the corresponding federated learning task to the terminal according to the federated learning task allocation result.

[0206] In an embodiment of the present disclosure, information related to a federated learning task is received from a wireless intelligent controller. If the information related to the federated learning task includes a federated learning task corresponding to the base station, the base station executes the federated learning task. If the information related to the federated learning task includes a federated learning task allocation strategy, the base station allocates a federated learning task result corresponding to the terminal according to the federated learning task allocation strategy, sends the federated learning task to the terminal according to the above-mentioned federated learning task allocation result, and feeds back the federated learning task allocation result to the wireless intelligent controller. The base station performs federated learning and feedback according to the instructions of the wireless intelligent controller and is intelligently managed by the wireless intelligent controller.

[0207] Optionally, after executing the federated learning task, the method further includes:

[0208] Sending learning indicator information to the wireless intelligent controller, where the learning indicator information is obtained by executing the federated learning task.

[0209] In an embodiment of the present disclosure, after executing the federated learning task, the base station sends learning indicator information to the wireless intelligent controller, so that the wireless intelligent controller monitors the status of the base station executing the federated learning task, and the learning indicator information includes the AI ​​service quality (Quality of Service, QoS) and / or key performance indicator (Key Performance Indicator, KPI) of the base station for federated learning, specifically, including the power (energy) and computing power consumed by training, computing time, federated learning model size, model accuracy, training speed, feedback cycle and other information.

[0210] Optionally, after sending the corresponding federated learning task to the terminal according to the federated learning task allocation result, the method further includes:

[0211] Obtaining learning indicator information corresponding to the execution of the federated learning task by the terminal;

[0212] The learning indicator information is sent to the wireless intelligent controller.

[0213] In an embodiment of the present disclosure, after sending the corresponding federated learning task to the terminal, the base station obtains learning indicator information of the terminal performing the task and sends it to the wireless intelligent controller. The learning indicator information includes the AI ​​service quality (QoS) and / or key performance indicator (KPI) of the terminal performing federated learning, specifically, including the power (energy) and computing power consumed by training, computing time, federated learning model size, model accuracy, training speed, feedback cycle and other information.

[0214] Optionally, before sending the learning indicator information to the wireless intelligent controller, the method further includes:

[0215] receiving a subscription request message sent by the wireless intelligent controller, where the subscription request message is used to request acquisition of the learning indicator information;

[0216] A subscription response message is sent to the wireless intelligent controller according to the subscription request message, where the subscription response message is used to confirm reporting of the learning indicator information.

[0217] In the disclosed embodiment, learning indicator information is reported through subscription by the wireless intelligent controller. After receiving the subscription request message, if the subscription request is agreed, a subscription response message is sent to the wireless intelligent controller. It should be noted that if the terminal participates in the execution of the federated learning task, the base station obtains the learning indicator information of the terminal in real time.

[0218] Optionally, the subscription request message carries a reporting condition for learning indicator information;

[0219] The learning indicator information is sent to the wireless intelligent controller according to the reporting condition.

[0220] In an embodiment of the present disclosure, the base station collects learning indicator information of the terminal in real time, and the reporting conditions include periodic reporting or quantity reporting. When the reporting condition is periodic reporting, the base station reports all the collected learning indicator information to the wireless intelligent controller in each period. At the same time, for the information of the terminal that is not reported, the base station can indicate the reason for non-reporting, including but not limited to inability to obtain, incomplete training task, etc.; when the reporting condition is quantity reporting, after collecting the learning indicator information of the terminals that meet the quantity, the base station reports the collected learning indicator information to the wireless intelligent controller, where the number of terminals in the quantity report can be an absolute number or a certain percentage of terminals.

[0221] Optionally, after sending the learning indicator information to the wireless intelligent controller, the method further includes:

[0222] Receiving an adjustment strategy corresponding to the federated learning task sent by the wireless intelligent controller;

[0223] Adjustments are made according to the adjustment strategy corresponding to the federated learning task.

[0224] In an embodiment of the present disclosure, the base station adjusts the federated learning task according to the adjustment strategy, or the base station adjusts the terminal side according to the adjustment strategy, wherein the adjustment strategy corresponding to the federated learning task includes one or more of the following: increasing the federated learning task, reducing the federated learning task, increasing the first device that executes the federated learning task, and reducing the first device that executes the federated learning task. The base station makes adjustments according to the adjustment strategy. When the adjustment strategy is a terminal-related adjustment strategy, the terminal and the corresponding federated learning task are adjusted according to the terminal-related adjustment strategy.

[0225] Optionally, the method further includes:

[0226] Sending first eigenvalue information to the wireless intelligent controller, where the first eigenvalue information includes a sub-eigenvalue of a model when the base station performs the federated learning task;

[0227] receiving first indication information fed back by the wireless intelligent controller according to the first characteristic value information;

[0228] The sub-feature values ​​of the model are updated according to the first indication information.

[0229] In an embodiment of the present disclosure, a method for updating a model is described. When executing the federated learning task, the base station sends a sub-eigenvalue of the model to the wireless intelligent controller. After receiving first indication information fed back by the wireless intelligent controller, the sub-eigenvalue is updated according to the first indication information. The first indication information includes the base station to be updated and the corresponding updated sub-eigenvalue. It should be noted that when the base station updates the sub-eigenvalue, it updates the corresponding training model, and then sends the updated training model to the wireless intelligent controller.

[0230] Optionally, after sending the corresponding federated learning task to the terminal according to the federated learning task allocation result, the method further includes:

[0231] Obtaining a sub-eigenvalue of a model sent by the terminal when the terminal executes the corresponding federated learning task;

[0232] Sending first eigenvalue information to the wireless intelligent controller, where the first eigenvalue information includes a sub-eigenvalue of a model when the terminal performs the federated learning task;

[0233] receiving first indication information fed back by the wireless intelligent controller according to the first characteristic value information;

[0234] Update information is sent to the terminal according to the first indication information, where the update information is used to instruct the terminal to update the sub-feature value of the model.

[0235] In an embodiment of the present disclosure, a method for updating a model is described. After the base station assigns a federated learning task to a terminal, the base station obtains the sub-eigenvalue of the model when the terminal performs the federated learning task, and sends the sub-eigenvalue of the terminal to the wireless intelligent controller. After receiving first indication information fed back by the wireless intelligent controller, update information is sent to the terminal according to the first indication information, instructing the terminal to update the sub-eigenvalue. The first indication information includes the terminal to be updated and the corresponding updated sub-eigenvalue. It should be noted that when the terminal updates the sub-eigenvalue, the corresponding training model is updated.

[0236] Optionally, after sending the update information to the terminal according to the first indication information, the method further includes:

[0237] Get the training model of the terminal;

[0238] Aggregating the training models of the terminals to obtain an aggregated model;

[0239] The aggregation model is sent to the wireless intelligent controller.

[0240] In the embodiment of the present disclosure, after the update information is sent to the terminal, the terminal will update the corresponding training model. The base station obtains the updated training model of the terminal and aggregates it to generate an aggregated model, and sends the aggregated model to the wireless intelligent controller, so that the wireless intelligent controller updates the model and even updates the information related to the federated learning task.

[0241] As shown in FIG3 , an embodiment of the present disclosure further provides a federated learning management method, which is applied to a terminal and includes:

[0242] Step S301: receiving federated learning task related information sent by a wireless intelligent controller through a base station, wherein the federated learning task related information includes a federated learning task, or receiving a federated learning task sent by the base station;

[0243] Step S302: Execute the federated learning task.

[0244] In an embodiment of the present disclosure, a federated learning task sent by a wireless intelligent controller is received by a base station, or a federated learning task sent by a base station is received, wherein the federated learning task sent by the base station is determined by the base station according to the federated learning task allocation strategy sent by the wireless intelligent controller, and the learning task is intelligently managed and allocated by the wireless intelligent controller.

[0245] Optionally, after executing the federated learning task, the method further includes:

[0246] Sending learning indicator information to the wireless intelligent controller through a base station or sending the learning indicator information to the base station;

[0247] The learning indicator information is obtained by executing the federated learning task.

[0248] In an embodiment of the present disclosure, after executing the federated learning task, learning indicator information is sent to the wireless intelligent controller or the base station, and the base station then feeds back the learning indicator information to the wireless intelligent controller, so that the wireless intelligent controller monitors the status of the terminal executing the federated learning task. The learning indicator information includes the AI ​​service quality (QoS) and / or key performance indicator (KPI) of the terminal performing federated learning, specifically, including the power (energy) and computing power consumed by training, computing time, federated learning model size, model accuracy, training speed, feedback cycle, and other information.

[0249] Optionally, the method further includes:

[0250] Sending the sub-eigenvalues ​​of the model when executing the federated learning task to the base station;

[0251] Obtaining update information sent by the base station;

[0252] The sub-feature values ​​of the model are updated according to the update information.

[0253] In an embodiment of the present disclosure, a method for updating the model is described. The sub-eigenvalues ​​of the model when executing the federated learning task are sent to the base station. After obtaining the update information sent by the base station, the sub-eigenvalues ​​are updated according to the new information. It should be noted that when the terminal updates the sub-eigenvalues, the corresponding training model is updated.

[0254] The federated learning management method provided in the present disclosure is applied to the O-RAN architecture. According to the business and intelligent application characteristics, wireless intelligent applications can be distributedly deployed. For example, rApp is deployed in a non-real-time intelligent controller to monitor the efficiency of federated training and model convergence, and an inference module is deployed in a near-real-time intelligent controller and / or a base station side and / or a terminal side. The near-real-time intelligent controller and / or the base station side and / or the terminal side perform inference on different layers. For example, the near-real-time intelligent controller is responsible for inference of cell-level configuration / policy, and the base station side is responsible for inference of terminal / link configuration / policy.

[0255] In summary, under the O-RAN intelligent architecture, the wireless intelligent controller can manage the federated learning of base stations and / or terminals, allocate and manage federated learning tasks, and manage inference tasks, thus realizing the application of federated learning in wireless networks. The wireless intelligent controller can also monitor the effects of federated learning in real time and optimize federated learning based on the monitored effects, thereby improving the computing resource utilization efficiency of terminals and / or base stations.

[0256] The present disclosure is further described below through several specific embodiments:

[0257] Example 1: As shown in FIG4 , the process of the wireless intelligent controller implementing federated learning task allocation and supervision of the terminal through the base station is as follows:

[0258] Step S401: The wireless intelligent controller obtains computing power information and / or attribute information of the terminal through subscription of the management service module;

[0259] Step S402: The wireless intelligent controller sends federated learning task related information to the base station, where the federated learning task related information includes the federated learning task allocation strategy of the terminal and the total federated learning task;

[0260] Step S403: The base station determines a task allocation result based on the federated learning task allocation strategy and the total federated learning tasks, where the task allocation result includes the terminals that will perform the federated learning tasks and the federated learning tasks corresponding to each terminal, and sends the corresponding federated learning tasks to the terminals.

[0261] Step S404: The base station feeds back the task allocation result to the wireless intelligent controller;

[0262] Step S405: The wireless intelligent controller sends a subscription request message to the base station, where the subscription request message is used to request to obtain the learning indicator information;

[0263] Step S406: After the base station agrees to the subscription request, it sends a subscription response message to the wireless intelligent controller;

[0264] Step S407: The base station obtains the learning indicator information of the terminal. It should be noted that the base station may obtain the learning indicator information of the terminal after sending the subscription response message, or may obtain the learning indicator information of the terminal in real time after step S403.

[0265] Step S408: Based on the subscription request message, the base station sends the learning indicator information of the terminal to the wireless intelligent controller. The subscription request message may carry a reporting condition, such as periodic reporting or quantity reporting. If the subscription request message carries the reporting condition, the base station sends the learning indicator information of the terminal to the wireless intelligent controller according to the reporting condition.

[0266] Example 2: As shown in FIG5 , after the steps of Example 1, the wireless intelligent controller can further adjust the federated learning task of the terminal through the base station. The specific process is as follows:

[0267] In step S501, the wireless intelligent controller sends an adjustment strategy corresponding to the federated learning task to the base station when the learning index information meets the first preset condition and / or the terminal computing power information is updated and / or the terminal attribute information is updated, wherein the terminal computing power information and / or the terminal attribute information can be obtained by subscription through the management service module, and can also be obtained by subscription through the base station or the terminal.

[0268] In step S502, the base station determines the federated learning task allocation strategy corresponding to the terminal according to the adjustment strategy corresponding to the federated learning task, and sends the learning task to the terminal according to the federated learning task allocation strategy corresponding to the terminal.

[0269] Example 3: As shown in FIG6 , during the process of the terminal executing the federated learning task, the wireless intelligent controller can also implement the update of the joint training model. The specific process is as follows:

[0270] Step S601: The base station obtains the sub-feature value of the terminal;

[0271] Step S602: The base station sends first characteristic value information to the wireless intelligent controller, where the first characteristic value information includes a terminal identifier and a corresponding sub-characteristic value.

[0272] In step S603, when the wireless intelligent controller determines that the sub-eigenvalue of at least one terminal needs to be updated, the wireless intelligent controller sends first indication information to the base station, where the first indication information includes the terminal that needs to be updated and the updated sub-eigenvalue corresponding to the terminal; it should be noted that when the wireless intelligent controller determines whether the sub-eigenvalue of the terminal needs to be updated, it can refer to the learning indicator information for judgment.

[0273] Step S604: The base station sends the updated sub-feature value to the terminal that needs to be updated according to the first indication information;

[0274] Step S605: The terminal updates the training model according to the updated sub-feature value;

[0275] Step S606: The terminal sends the updated training model to the base station;

[0276] Step S607: The base station aggregates the received training models to obtain an aggregated model;

[0277] Step S608: The base station sends the aggregation model to the wireless intelligent controller.

[0278] In addition, after step S608, the wireless intelligent controller may also update information related to the federated learning task according to the aggregation model, such as federated learning content or a federated learning allocation strategy of the terminal.

[0279] As shown in FIG7 , an embodiment of the present disclosure further provides a network device 700, including a transceiver 701 and a processor 702, wherein the transceiver 701 is configured to:

[0280] Obtain computing power information and / or attribute information of a first device, where the first device includes a base station and / or a terminal;

[0281] According to the computing power information and / or attribute information, federated learning task related information is sent to the first device, where the federated learning task related information includes a federated learning task, or the federated learning task related information is used to determine the federated learning task.

[0282] Optionally, the terminal attribute information acquired by the transceiver 701 includes one or more of the following:

[0283] Terminal location information, terminal type, service type, and service priority.

[0284] Optionally, the processor 702 is configured to: determine, based on the computing power information and / or the attribute information, information related to a federated learning task corresponding to the terminal, where the information related to the federated learning task corresponding to the terminal includes a federated learning task allocation strategy, and the federated learning task allocation strategy is used to determine the federated learning task;

[0285] The transceiver 701 is further configured to send information related to the federated learning task corresponding to the terminal to the base station.

[0286] Optionally, the transceiver 701 is further configured to:

[0287] Receive a federated learning task allocation result sent by the base station, where the federated learning task allocation result is determined according to the federated learning task allocation strategy, and the federated learning task allocation result includes: a terminal that executes a federated learning task and a federated learning task corresponding to each terminal.

[0288] Optionally, the processor 702 is configured to: determine, based on computing power information and / or attribute information of the terminal, information related to a federated learning task corresponding to the terminal, where the information related to the federated learning task corresponding to the terminal includes: a terminal executing the federated learning task and a federated learning task corresponding to each terminal;

[0289] The transceiver 701 is further configured to send information related to a federated learning task corresponding to the terminal to the terminal through the base station.

[0290] Optionally, the processor 702 is configured to: determine, based on computing power information and / or attribute information of the base station, information related to a federated learning task corresponding to the base station, where the information related to the federated learning task corresponding to the base station includes: a base station that performs the federated learning task and a federated learning task corresponding to each base station;

[0291] The transceiver 701 is further configured to send information related to the federated learning task corresponding to the base station to the base station.

[0292] Optionally, the transceiver 701 is further configured to:

[0293] Obtaining learning indicator information sent by the first device, where the learning indicator information is obtained by the first device performing the federated learning task;

[0294] According to the learning indicator information, an adjustment strategy corresponding to the federated learning task is sent to the first device, or a model is assigned to the inference model obtained by the first device executing the federated learning task.

[0295] Optionally, the transceiver 701 is further configured to:

[0296] Sending a subscription request message to the base station, where the subscription request message is used to request obtaining the learning indicator information;

[0297] A subscription response message is received from the base station according to the subscription request message, where the subscription response message is used to confirm reporting of the learning indicator information.

[0298] Optionally, the subscription request message sent by the transceiver 701 carries a reporting condition for the learning indicator information;

[0299] The transceiver 701 is further configured to obtain the learning indicator information sent by the base station according to the reporting condition.

[0300] Optionally, the processor 702 is configured to: determine an adjustment strategy corresponding to the federated learning task when it is determined that the learning indicator information meets a first preset condition;

[0301] The transceiver 701 is further configured to send an adjustment strategy corresponding to the federated learning task to the first device.

[0302] Optionally, the adjustment strategy corresponding to the federated learning task sent by the transceiver 701 includes one or more of the following:

[0303] Increase federated learning tasks, reduce federated learning tasks, increase the first device that executes federated learning tasks, and reduce the first device that executes federated learning tasks.

[0304] Optionally, the learning indicator information received by the transceiver 701 includes one or more of the following:

[0305] Model training speed, computing power consumption, and model training effect;

[0306] The first preset condition used by the processor 702 to determine includes:

[0307] The model training speed is less than a first threshold and / or the consumed computing power is greater than a second threshold and / or the model training effect does not meet the preset effect.

[0308] Optionally, the processor 702 is configured to: if it is determined that the learning indicator information satisfies a second preset condition, determine the currently obtained aggregate model as an inference model, where the aggregate model is obtained by aggregating models generated by the first device when performing the federated learning task;

[0309] The transceiver 701 is used to: distribute the inference model.

[0310] Optionally, the transceiver 701 is configured to: obtain task requirement information of a service to be processed, wherein the task requirement information includes real-time requirements and / or input and output requirements;

[0311] The transceiver 701 is used to: distribute the reasoning model according to the task requirement information.

[0312] Optionally, the transceiver 701 is configured to:

[0313] Obtaining first eigenvalue information sent by the first device, where the first eigenvalue information includes sub-eigenvalues ​​of a model when the first device performs the federated learning task;

[0314] According to the first eigenvalue information, first indication information is sent to the first device, where the first indication information is used to instruct the first device to update the sub-eigenvalue of the model.

[0315] It should be noted that the device in this embodiment corresponds to the method applied to the wireless intelligent controller. The implementation methods in the above embodiments are applicable to the embodiments of this device and can achieve the same technical effects. The above device provided in the embodiments of the present disclosure can implement all the method steps implemented in the above method embodiments and can achieve the same technical effects. The parts and beneficial effects of this embodiment that are the same as those in the method embodiments will not be described in detail here.

[0316] As shown in FIG8 , an embodiment of the present disclosure further provides a base station 800, including a transceiver 801 and a processor 802, wherein:

[0317] The transceiver 801 is configured to receive federated learning task related information sent by a wireless intelligent controller, where the federated learning task related information includes a federated learning task or a federated learning task allocation strategy;

[0318] The processor 802 is configured to: execute the federated learning task if the federated learning task related information includes the federated learning task;

[0319] The processor 802 is configured to: when the federated learning task related information includes the federated learning task allocation strategy, determine a federated learning task allocation result corresponding to the terminal according to the federated learning task allocation strategy, where the federated learning task allocation result includes a terminal that performs the federated learning task and a federated learning task corresponding to each terminal;

[0320] The transceiver 801 is configured to: send the federated learning task allocation result to the wireless intelligent controller, and send the corresponding federated learning task to the terminal according to the federated learning task allocation result.

[0321] Optionally, the transceiver 801 is further configured to:

[0322] Sending learning indicator information to the wireless intelligent controller, where the learning indicator information is obtained by executing the federated learning task.

[0323] Optionally, the transceiver 801 is further configured to:

[0324] Obtaining learning indicator information corresponding to the execution of the federated learning task by the terminal;

[0325] The learning indicator information is sent to the wireless intelligent controller.

[0326] Optionally, the transceiver 801 is further configured to:

[0327] receiving a subscription request message sent by the wireless intelligent controller, where the subscription request message is used to request acquisition of the learning indicator information;

[0328] A subscription response message is sent to the wireless intelligent controller according to the subscription request message, where the subscription response message is used to confirm reporting of the learning indicator information.

[0329] Optionally, the subscription request message obtained by the transceiver 801 carries a reporting condition for the learning indicator information;

[0330] The transceiver 801 is further configured to:

[0331] The learning indicator information is sent to the wireless intelligent controller according to the reporting condition.

[0332] Optionally, the transceiver 801 is further configured to receive an adjustment strategy corresponding to the federated learning task sent by the wireless intelligent controller;

[0333] The processor 802 is further configured to perform adjustments according to an adjustment strategy corresponding to the federated learning task.

[0334] Optionally, the transceiver 801 is further configured to send first eigenvalue information to the wireless intelligent controller, where the first eigenvalue information includes a sub-eigenvalue of a model when the base station performs the federated learning task;

[0335] The transceiver 801 is further configured to receive first indication information fed back by the wireless intelligent controller according to the first characteristic value information;

[0336] The processor 802 is further configured to update the sub-feature value of the model according to the first indication information.

[0337] Optionally, the transceiver 801 is further configured to:

[0338] Obtaining a sub-eigenvalue of a model sent by the terminal when the terminal executes the corresponding federated learning task;

[0339] Sending first eigenvalue information to the wireless intelligent controller, where the first eigenvalue information includes a sub-eigenvalue of a model when the terminal performs the federated learning task;

[0340] receiving first indication information fed back by the wireless intelligent controller according to the first characteristic value information;

[0341] Update information is sent to the terminal according to the first indication information, where the update information is used to instruct the terminal to update the sub-feature value of the model.

[0342] It should be noted that the device in this embodiment is a device corresponding to the method applied to the base station side, and the implementation methods in the above embodiments are all applicable to the embodiments of this device and can achieve the same technical effects. The above-mentioned device provided in the embodiment of the present disclosure can implement all the method steps implemented in the above-mentioned method embodiment and can achieve the same technical effects. The parts and beneficial effects of this embodiment that are the same as those in the method embodiment will not be described in detail here.

[0343] As shown in FIG9 , an embodiment of the present disclosure further provides a terminal, including a transceiver 901 and a processor 902, wherein:

[0344] The transceiver 901 is configured to: receive federated learning task related information sent by a wireless intelligent controller through a base station, wherein the federated learning task related information includes a federated learning task, or receive a federated learning task sent by the base station;

[0345] The processor 902 is configured to execute the federated learning task.

[0346] Optionally, the transceiver 901 is also used to

[0347] The learning indicator information is sent to the wireless intelligent controller through the base station or the learning indicator information is sent to the base station; wherein the learning indicator information is obtained by executing the federated learning task.

[0348] Optionally, the transceiver 901 is configured to send a sub-eigenvalue of the model when executing the federated learning task to the base station;

[0349] The transceiver 901 is used to obtain the update information sent by the base station;

[0350] The processor 902 is configured to update the sub-feature value of the model according to the update information.

[0351] It should be noted that the device in this embodiment is a device corresponding to the method applied to the terminal side, and the implementation methods in the above embodiments are all applicable to the embodiments of the device and can achieve the same technical effects. The above-mentioned device provided by the embodiment of the present disclosure can implement all the method steps implemented by the above-mentioned method embodiment and can achieve the same technical effects. The parts and beneficial effects of this embodiment that are the same as those in the method embodiment will not be described in detail here.

[0352] As shown in FIG10 , an embodiment of the present disclosure further provides a federated learning management method and apparatus, which is applied to an intelligent controller and includes:

[0353] A first acquisition module 1001 is configured to acquire computing power information and / or attribute information of a first device, where the first device includes a base station and / or a terminal;

[0354] The first sending module 1002 is used to send federated learning task related information to the first device based on the computing power information and / or attribute information, where the federated learning task related information includes the federated learning task, or the federated learning task related information is used to determine the federated learning task.

[0355] Optionally, the terminal attribute information acquired by the first acquiring module 1001 includes one or more of the following:

[0356] Terminal location information, terminal type, service type, and service priority.

[0357] Optionally, the first sending module 1002 includes:

[0358] a first determining unit, configured to determine, based on the computing power information and / or the attribute information, information related to a federated learning task corresponding to a terminal, wherein the information related to the federated learning task corresponding to the terminal includes a federated learning task allocation strategy, and the federated learning task allocation strategy is used to determine the federated learning task;

[0359] The first sending unit is configured to send, by the base station, information related to the federated learning task corresponding to the terminal.

[0360] Optionally, the device further comprises:

[0361] The third receiving module is used to receive the federated learning task allocation result sent by the base station, where the federated learning task allocation result is determined according to the federated learning task allocation strategy, and the federated learning task allocation result includes: the terminal that executes the federated learning task and the federated learning task corresponding to each terminal.

[0362] Optionally, the first sending module 1002 further includes:

[0363] a second determining unit, configured to determine, based on the computing power information and / or attribute information of the terminal, information related to the federated learning task corresponding to the terminal, wherein the information related to the federated learning task corresponding to the terminal includes: the terminal executing the federated learning task and the federated learning task corresponding to each terminal;

[0364] The second sending unit is configured to send information related to the federated learning task corresponding to the terminal to the terminal through the base station.

[0365] Optionally, the first sending module 1002 further includes:

[0366] a third determining unit, configured to determine, based on the computing power information and / or attribute information of the base station, information related to the federated learning task corresponding to the base station, where the information related to the federated learning task corresponding to the base station includes: the base station that performs the federated learning task and the federated learning task corresponding to each base station;

[0367] The third sending unit is used to send information related to the federated learning task corresponding to the base station to the base station.

[0368] Optionally, the device further comprises:

[0369] A second acquisition module is configured to acquire learning indicator information sent by the first device, where the learning indicator information is obtained by the first device performing the federated learning task;

[0370] The third sending module is used to send the adjustment strategy corresponding to the federated learning task to the first device based on the learning indicator information, or to perform model allocation on the inference model obtained by the first device executing the federated learning task.

[0371] Optionally, the device further comprises:

[0372] A fourth sending module is used to send a subscription request message to the base station, where the subscription request message is used to request to obtain the learning indicator information;

[0373] A fourth receiving module is used to receive a subscription response message sent by the base station according to the subscription request message, where the subscription response message is used to confirm reporting of the learning indicator information.

[0374] Optionally, the subscription request message sent by the fourth sending module carries a reporting condition for the learning indicator information;

[0375] The second acquisition module includes:

[0376] The first acquisition unit is configured to acquire the learning indicator information sent by the base station according to the reporting condition.

[0377] Optionally, the third sending module includes:

[0378] a fourth determining unit, configured to determine an adjustment strategy corresponding to the federated learning task when it is determined that the learning indicator information satisfies the first preset condition;

[0379] The fourth sending unit is used to send the adjustment strategy corresponding to the federated learning task to the first device.

[0380] Optionally, the adjustment strategy corresponding to the federated learning task sent by the third sending module includes one or more of the following:

[0381] Increase federated learning tasks, reduce federated learning tasks, increase the first device that executes federated learning tasks, and reduce the first device that executes federated learning tasks.

[0382] Optionally, the learning indicator information in the fourth determining unit includes one or more of the following:

[0383] Model training speed, computing power consumption, and model training effect;

[0384] The first preset condition in the fourth determining unit includes:

[0385] The model training speed is less than a first threshold and / or the consumed computing power is greater than a second threshold and / or the model training effect does not meet the preset effect.

[0386] Optionally, the third sending module includes:

[0387] a fifth determining unit, configured to, if it is determined that the learning indicator information satisfies a second preset condition, determine the currently obtained aggregate model as an inference model, where the aggregate model is obtained by aggregating models generated by the first device when executing the federated learning task;

[0388] The first allocating unit is configured to perform model allocation on the inference model.

[0389] Optionally, the first allocating unit includes:

[0390] A second acquiring unit is configured to acquire task requirement information of a business to be processed, wherein the task requirement information includes real-time requirements and / or input and output requirements;

[0391] The second allocating unit is configured to allocate the inference model according to the task requirement information.

[0392] Optionally, the device further comprises:

[0393] a third acquisition module, configured to acquire first eigenvalue information sent by the first device, where the first eigenvalue information includes a sub-eigenvalue of a model when the first device performs the federated learning task;

[0394] The fifth sending module is used to send first indication information to the first device according to the first eigenvalue information, where the first indication information is used to instruct the first device to update the sub-eigenvalue of the model.

[0395] It should be noted that the embodiment of the device is a device corresponding to the embodiment of the above method, and all implementation methods in the embodiment of the above method are applicable to the embodiment of the device and can achieve the same technical effect.

[0396] As shown in FIG11 , an embodiment of the present disclosure further provides a federated learning management method and apparatus, which is applied to a base station and includes:

[0397] A first receiving module 1101 is configured to receive federated learning task related information sent by a wireless intelligent controller, where the federated learning task related information includes a federated learning task or a federated learning task allocation strategy;

[0398] A first execution module 1102 is configured to execute the federated learning task if the federated learning task related information includes the federated learning task;

[0399] A first determining module 1103 is configured to, when the federated learning task related information includes the federated learning task allocation policy, determine a federated learning task allocation result corresponding to the terminal according to the federated learning task allocation policy, wherein the federated learning task allocation result includes a terminal that performs the federated learning task and a federated learning task corresponding to each terminal;

[0400] The second sending module 1104 is configured to send the federated learning task allocation result to the wireless intelligent controller, and send the corresponding federated learning task to the terminal according to the federated learning task allocation result.

[0401] Optionally, the device further comprises:

[0402] A sixth sending module is used to send learning indicator information to the wireless intelligent controller, where the learning indicator information is obtained by executing the federated learning task.

[0403] Optionally, the device further comprises:

[0404] A fourth acquisition module is used to obtain learning indicator information corresponding to the execution of the federated learning task by the terminal;

[0405] A seventh sending module is configured to send the learning indicator information to the wireless intelligent controller.

[0406] Optionally, the device further comprises:

[0407] a fifth receiving module, configured to receive a subscription request message sent by the wireless intelligent controller, wherein the subscription request message is used to request acquisition of the learning indicator information;

[0408] An eighth sending module is configured to send a subscription response message to the wireless intelligent controller according to the subscription request message, wherein the subscription response message is used to confirm reporting of the learning indicator information.

[0409] Optionally, the subscription request message received by the fifth receiving module carries a reporting condition for the learning indicator information;

[0410] The seventh sending module includes:

[0411] A fifth sending unit is configured to send learning indicator information to the wireless intelligent controller according to the reporting condition.

[0412] Optionally, the device further comprises:

[0413] a sixth receiving module, configured to receive an adjustment strategy corresponding to the federated learning task sent by the wireless intelligent controller;

[0414] The first adjustment module is used to make adjustments according to the adjustment strategy corresponding to the federated learning task.

[0415] Optionally, the device further comprises:

[0416] an eighth sending module, configured to send first eigenvalue information to the wireless intelligent controller, where the first eigenvalue information includes a sub-eigenvalue of a model when the base station performs the federated learning task;

[0417] A seventh receiving module, configured to receive first indication information fed back by the wireless intelligent controller according to the first characteristic value information;

[0418] A first updating module is configured to update the sub-feature value of the model according to the first indication information.

[0419] Optionally, the device further comprises:

[0420] a fifth acquisition module, configured to acquire a sub-eigenvalue of the model sent by the terminal when the terminal performs the corresponding federated learning task;

[0421] a ninth sending module, configured to send first eigenvalue information to the wireless intelligent controller, where the first eigenvalue information includes a sub-eigenvalue of the model when the terminal performs the federated learning task;

[0422] An eighth receiving module, configured to receive first indication information fed back by the wireless intelligent controller according to the first characteristic value information;

[0423] A tenth sending module is used to send update information to the terminal according to the first indication information, where the update information is used to instruct the terminal to update the sub-feature value of the model.

[0424] It should be noted that the embodiment of the device is a device corresponding to the embodiment of the above method, and all implementation methods in the embodiment of the above method are applicable to the embodiment of the device and can achieve the same technical effect.

[0425] As shown in FIG12 , an embodiment of the present disclosure further provides a federated learning management method and apparatus, which is applied to a terminal and includes:

[0426] A second receiving module 1201 is configured to receive federated learning task related information sent by a wireless intelligent controller through a base station, wherein the federated learning task related information includes a federated learning task, or receive a federated learning task sent by the base station;

[0427] The second execution module 1202 is used to execute the federated learning task.

[0428] Optionally, the device further comprises:

[0429] An eleventh sending module is configured to send learning indicator information to the wireless intelligent controller via a base station or to send the learning indicator information to the base station; wherein the learning indicator information is obtained by executing the federated learning task.

[0430] Optionally, the device further comprises:

[0431] a twelfth sending module, configured to send the sub-eigenvalues ​​of the model when executing the federated learning task to the base station;

[0432] A sixth acquisition module, configured to acquire the update information sent by the base station;

[0433] The second updating module is used to update the sub-feature values ​​of the model according to the update information.

[0434] It should be noted that the embodiment of the device is a device corresponding to the embodiment of the above method, and all implementation methods in the embodiment of the above method are applicable to the embodiment of the device and can achieve the same technical effect.

[0435] An embodiment of the present disclosure further provides a network device, comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the federated learning management method as described in any one of the above items.

[0436] An embodiment of the present disclosure further provides a readable storage medium, including: a program stored on the readable storage medium, and when the program is executed by a processor, the steps of the federated learning management method as described in any one of the above items are implemented.

[0437] An embodiment of the present disclosure further provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the steps of the federated learning management method as described in any one of the above items.

[0438] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0439] Obviously, those skilled in the art may make various changes and modifications to the present disclosure without departing from the spirit and scope of the present disclosure. Thus, if these modifications and variations of the present disclosure fall within the scope of the claims of the present disclosure and their equivalents, the present disclosure is intended to include these modifications and variations.

Claims

1. A federated learning management method, applied to a wireless intelligent controller, comprising: Obtain computing power information and / or attribute information of a first device, where the first device includes a base station and / or a terminal; According to the computing power information and / or attribute information, federated learning task related information is sent to the first device, where the federated learning task related information includes a federated learning task, or the federated learning task related information is used to determine the federated learning task.

2. The federated learning management method according to claim 1, wherein: The terminal attribute information includes one or more of the following: Terminal location information, terminal type, service type, and service priority.

3. The federated learning management method according to claim 1, wherein: The sending, according to the computing power information and / or attribute information, the federated learning task related information to the first device includes: Determining, based on the computing power information and / or the attribute information, information related to a federated learning task corresponding to the terminal, where the information related to the federated learning task corresponding to the terminal includes a federated learning task allocation strategy, where the federated learning task allocation strategy is used to determine the federated learning task; Sending information related to the federated learning task corresponding to the terminal to the base station.

4. The federated learning management method according to claim 3, wherein: After sending the federated learning task related information corresponding to the terminal to the base station, the method further includes: Receive a federated learning task allocation result sent by the base station, where the federated learning task allocation result is determined according to the federated learning task allocation strategy, and the federated learning task allocation result includes: a terminal that executes a federated learning task and a federated learning task corresponding to each terminal.

5. The federated learning management method according to claim 1, wherein: The sending, according to the computing power information and / or attribute information, the federated learning task related information to the first device includes: Determine, based on the computing power information and / or attribute information of the terminal, information related to the federated learning task corresponding to the terminal, where the information related to the federated learning task corresponding to the terminal includes: the terminal executing the federated learning task and the federated learning task corresponding to each terminal; The federated learning task related information corresponding to the terminal is sent to the terminal through the base station.

6. The federated learning management method according to claim 1, wherein: The sending, according to the computing power information and / or attribute information, the federated learning task related information to the first device includes: Determine, based on the computing power information and / or attribute information of the base station, information related to the federated learning task corresponding to the base station, where the information related to the federated learning task corresponding to the base station includes: the base station that performs the federated learning task and the federated learning task corresponding to each base station; Sending information related to the federated learning task corresponding to the base station to the base station.

7. The federated learning management method according to claim 1, wherein: After sending the federated learning task related information to the first device, the method further includes: Obtaining learning indicator information sent by the first device, where the learning indicator information is obtained by the first device performing the federated learning task; According to the learning indicator information, an adjustment strategy corresponding to the federated learning task is sent to the first device, or a model is assigned to the inference model obtained by the first device executing the federated learning task.

8. The federated learning management method according to claim 7, wherein: Before acquiring the learning indicator information sent by the first device, the method further includes: Sending a subscription request message to the base station, where the subscription request message is used to request obtaining the learning indicator information; A subscription response message is received from the base station according to the subscription request message, where the subscription response message is used to confirm reporting of the learning indicator information.

9. The federated learning management method according to claim 8, wherein: The subscription request message carries the reporting conditions of the learning indicator information; Obtaining learning indicator information sent by the first device, including: Obtain the learning indicator information sent by the base station according to the reporting condition.

10. The federated learning management method according to claim 7, wherein: The sending, according to the learning indicator information, an adjustment strategy corresponding to the federated learning task to the first device includes: When it is determined that the learning indicator information satisfies the first preset condition, determining an adjustment strategy corresponding to the federated learning task; Sending an adjustment strategy corresponding to the federated learning task to the first device.

11. The federated learning management method according to claim 7, wherein: The adjustment strategy corresponding to the federated learning task includes one or more of the following: Increase federated learning tasks, reduce federated learning tasks, increase the first device that executes federated learning tasks, and reduce the first device that executes learning tasks.

12. The federated learning management method according to claim 10, wherein: The learning indicator information includes one or more of the following: Model training speed, computing power consumption, and model training effect; The first preset condition includes: The model training speed is less than a first threshold and / or the consumed computing power is greater than a second threshold and / or the model training effect does not meet the preset effect.

13. The federated learning management method according to claim 7, wherein: The performing model allocation on the inference model obtained by the first device executing the federated learning task according to the learning indicator information includes: If it is determined that the learning indicator information satisfies a second preset condition, determining the currently obtained aggregate model as an inference model, where the aggregate model is obtained by aggregating models generated by the first device when executing the federated learning task; A model assignment is performed on the inference model.

14. The federated learning management method according to claim 13, wherein: Performing model assignment on the inference model includes: Obtaining task requirement information of the business to be processed, wherein the task requirement information includes real-time requirements and / or input and output requirements; The reasoning model is assigned a model according to the task requirement information.

15. The federated learning management method according to claim 1 or 7, wherein: After sending the federated learning task related information to the first device, the method further includes: Obtaining first eigenvalue information sent by the first device, where the first eigenvalue information includes sub-eigenvalues ​​of a model when the first device performs the federated learning task; According to the first eigenvalue information, first indication information is sent to the first device, where the first indication information is used to instruct the first device to update the sub-eigenvalue of the model.

16. A federated learning management method, applied to a base station, comprising: Receiving federated learning task related information sent by the wireless intelligent controller, wherein the federated learning task related information includes a federated learning task or a federated learning task allocation strategy; In a case where the federated learning task-related information includes the federated learning task, executing the federated learning task; In a case where the federated learning task related information includes the federated learning task allocation strategy, determining a federated learning task allocation result corresponding to the terminal according to the federated learning task allocation strategy, the federated learning task allocation result including the terminal that performs the federated learning task and the federated learning task corresponding to each terminal; The federated learning task allocation result is sent to the wireless intelligent controller, and the corresponding federated learning task is sent to the terminal according to the federated learning task allocation result.

17. The federated learning management method according to claim 16, wherein: After executing the federated learning task, the method further includes: Sending learning indicator information to the wireless intelligent controller, where the learning indicator information is obtained by executing the federated learning task.

18. The federated learning management method according to claim 16, wherein: After sending the corresponding federated learning task to the terminal according to the federated learning task allocation result, the method further includes: Obtaining learning indicator information corresponding to the execution of the federated learning task by the terminal; The learning indicator information is sent to the wireless intelligent controller.

19. The federated learning management method according to claim 17 or 18, wherein: Before sending the learning indicator information to the wireless intelligent controller, the method further includes: receiving a subscription request message sent by the wireless intelligent controller, where the subscription request message is used to request acquisition of the learning indicator information; A subscription response message is sent to the wireless intelligent controller according to the subscription request message, where the subscription response message is used to confirm reporting of the learning indicator information.

20. The federated learning management method according to claim 19, wherein: The subscription request message carries the reporting conditions of the learning indicator information; The sending the learning indicator information to the wireless intelligent controller includes: The learning indicator information is sent to the wireless intelligent controller according to the reporting condition.

21. The federated learning management method according to claim 17 or 18, wherein: After sending the learning indicator information to the wireless intelligent controller, the method further includes: Receiving an adjustment strategy corresponding to the federated learning task sent by the wireless intelligent controller; Adjustments are made according to the adjustment strategy corresponding to the federated learning task.

22. The federated learning management method according to claim 16, further comprising: Sending first eigenvalue information to the wireless intelligent controller, where the first eigenvalue information includes a sub-eigenvalue of a model when the base station performs the federated learning task; receiving first indication information fed back by the wireless intelligent controller according to the first characteristic value information; The sub-feature values ​​of the model are updated according to the first indication information.

23. The federated learning management method according to claim 16 or 22, wherein: After sending the corresponding federated learning task to the terminal according to the federated learning task allocation result, the method further includes: Obtaining a sub-eigenvalue of a model sent by the terminal when the terminal executes the corresponding federated learning task; Sending first eigenvalue information to the wireless intelligent controller, where the first eigenvalue information includes a sub-eigenvalue of a model when the terminal performs the federated learning task; receiving first indication information fed back by the wireless intelligent controller according to the first characteristic value information; Update information is sent to the terminal according to the first indication information, where the update information is used to instruct the terminal to update the sub-feature value of the model.

24. A federated learning management method, applied to a terminal, comprising: Receiving federated learning task related information sent by a wireless intelligent controller through a base station, wherein the federated learning task related information includes a federated learning task, or receiving a federated learning task sent by the base station; Execute the federated learning task.

25. The federated learning management method according to claim 24, wherein: After executing the federated learning task, the method further includes: Sending learning indicator information to the wireless intelligent controller through a base station or sending the learning indicator information to the base station; The learning indicator information is obtained by executing the federated learning task.

26. The federated learning management method according to claim 24, further comprising: Sending the sub-eigenvalues ​​of the model when executing the federated learning task to the base station; Obtaining update information sent by the base station; The sub-feature values ​​of the model are updated according to the update information.

27. A network device comprising a transceiver and a processor, wherein: The transceiver is used for: Obtain computing power information and / or attribute information of a first device, where the first device includes a base station and / or a terminal; According to the computing power information and / or attribute information, federated learning task related information is sent to the first device, where the federated learning task related information includes a federated learning task, or the federated learning task related information is used to determine the federated learning task.

28. A base station comprising a transceiver and a processor, wherein: The transceiver is used to: receive federated learning task related information sent by the wireless intelligent controller, wherein the federated learning task related information includes a federated learning task or a federated learning task allocation strategy; The processor is configured to: execute the federated learning task when the federated learning task related information includes the federated learning task; The processor is configured to: when the federated learning task related information includes the federated learning task allocation strategy, determine a federated learning task allocation result corresponding to the terminal according to the federated learning task allocation strategy, wherein the federated learning task allocation result includes a terminal that performs the federated learning task and a federated learning task corresponding to each terminal; The transceiver is used to: send the federated learning task allocation result to the wireless intelligent controller, and send the corresponding federated learning task to the terminal according to the federated learning task allocation result.

29. A terminal comprising a transceiver and a processor, wherein: The transceiver is configured to: receive federated learning task related information sent by a wireless intelligent controller through a base station, wherein the federated learning task related information includes a federated learning task, or receive a federated learning task sent by the base station; The processor is used to: execute the federated learning task.

30. A federated learning management method and apparatus, applied to an intelligent controller, comprising: A first acquisition module is configured to acquire computing power information and / or attribute information of a first device, where the first device includes a base station and / or a terminal; A first sending module is used to send federated learning task related information to the first device based on the computing power information and / or attribute information, where the federated learning task related information includes the federated learning task, or the federated learning task related information is used to determine the federated learning task.

31. A federated learning management method and apparatus, applied to a base station, comprising: A first receiving module is configured to receive federated learning task related information sent by the wireless intelligent controller, wherein the federated learning task related information includes a federated learning task or a federated learning task allocation strategy; a first execution module, configured to execute the federated learning task if the federated learning task related information includes the federated learning task; a first determining module configured to, when the federated learning task related information includes the federated learning task allocation strategy, determine a federated learning task allocation result corresponding to the terminal according to the federated learning task allocation strategy, the federated learning task allocation result including the terminal that performs the federated learning task and the federated learning task corresponding to each terminal; The second sending module is used to send the federated learning task allocation result to the wireless intelligent controller, and send the corresponding federated learning task to the terminal according to the federated learning task allocation result.

32. A federated learning management method and apparatus, applied to a terminal, comprising: A second receiving module is configured to receive federated learning task related information sent by the wireless intelligent controller through the base station, wherein the federated learning task related information includes the federated learning task, or receive the federated learning task sent by the base station; The second execution module is used to execute the federated learning task.

33. A network device comprising: A processor, a memory, and a program stored on the memory and executable on the processor, wherein the program, when executed by the processor, implements the federated learning management method according to any one of claims 1 to 15, or the federated learning management method according to any one of claims 16 to 23, or the federated learning management method according to any one of claims 24 to 26.

34. A readable storage medium comprising: The readable storage medium stores a program, and when the program is executed by the processor, it implements the steps of the federated learning management method according to any one of claims 1 to 15, or implements the steps of the federated learning management method according to any one of claims 16 to 23, or implements the steps of the federated learning management method according to any one of claims 24 to 26.

35. A computer program product comprising computer instructions, wherein when the computer instructions are executed by a processor, the steps of the federated learning management method according to any one of claims 1 to 15 are implemented, or the steps of the federated learning management method according to any one of claims 16 to 23 are implemented, or the steps of the federated learning management method according to any one of claims 24 to 26 are implemented.

Images