Federal learning method, federal learning server AI network element and communication system
By determining the federated learning strategy in federated learning and selecting the appropriate client for model training and aggregation, the problem of large differences in data characteristics of multiple clients is solved, and the accuracy and generalization of the model are improved.
Patent Information
- Application Number
- CN202411813157.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-10
- Publication Date
- 2025-03-18
AI Technical Summary
In federated learning, when the data characteristics of multiple clients vary greatly, the performance accuracy and generalization of the model are poor.
Determine the federated learning strategy through the federated learning server AI network element, select the appropriate federated learning client AI network element, and issue the federated learning global model to it for training, generate the local federated learning model, and perform aggregation.
It effectively improves the accuracy and generalization of the federated learning model, thereby improving the user experience.
Smart Images
Figure CN119647628A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of communications, and in particular to a federated learning method, a federated learning server AI (Artificial Intelligence) network element, and a communication system. Background Art
[0002] Federated learning is a distributed learning method that can complete model training while ensuring user data privacy by training locally on the federated learning client and aggregating models on the federated learning server. Currently, federated learning is one of the research focuses of 3GPP (3rd Generation Partnership Project) in network intelligence research. Summary of the invention
[0003] The inventors noted that in the related art, the federated learning client only needs to perform local training and upload the trained local model to the federated learning server without uploading any local data. However, when the data features between multiple clients are quite different (for example, data comes from different regions, different times, etc.), the data of each client is not independent and identically distributed, resulting in poor performance accuracy and generalization of the federated learning model.
[0004] Based on this, the present disclosure provides a federated learning method that can effectively improve the accuracy and generalization of the federated learning model, thereby improving the user experience.
[0005] In a first aspect of the present disclosure, a federated learning method is provided, which is executed by a federated learning server AI network element, including: determining a federated learning strategy according to a business requirement included in a model training request sent by a user device; using a registration network element to discover a candidate federated learning client AI network element; selecting a federated learning client AI network element from the candidate federated learning client AI network elements according to the federated learning strategy; issuing a federated learning global model to the federated learning client AI network element so that the federated learning client AI network element trains the federated learning global model to generate a local federated learning model; aggregating the local federated learning model uploaded by the federated learning client AI network element; and repeatedly performing the step of selecting a federated learning client AI network element until the federated learning is completed.
[0006] In some embodiments, the federated learning strategy includes data privacy level information of federated learning members, and the proportion of each data privacy level in the data privacy level information.
[0007] In some embodiments, the use of a registered network element to discover a candidate federated learning client AI network element includes: sending a federated learning client AI network element discovery request to the registered network element, wherein the discovery request includes the data privacy level information, so that the registered network element discovers the candidate federated learning client AI network element corresponding to the data privacy level information; and receiving the federated learning client AI network element discovery result sent by the registered network element.
[0008] In some embodiments, the local federated learning model is obtained by the federated learning client AI network element using a local data set to train the federated learning global model.
[0009] In some embodiments, a shared data set is sent to the federated learning client AI network element so that the federated learning client AI network element trains the federated learning global model using the shared data set and the local data set.
[0010] In some embodiments, the data uploaded by the federated learning client AI network element is integrated.
[0011] In some embodiments, a registration request is sent to the registration network element, wherein the registration request includes data privacy level information of the federated learning server AI network element.
[0012] In some embodiments, the data privacy level of the federated learning server-side AI network element includes one of the following: the first level is used to indicate support for direct collection of any data; the second level is used to indicate support for local training and uploading of the federated learning model; the third level is used to indicate support for local training and uploading of the federated learning model, and support for uploading of part of the data that will contribute to the training of the federated learning model, wherein the privacy of the part of the data is lower than the specified level, and the amount of the part of the data is less than a predetermined proportion; the fourth level is used to indicate support for local training and uploading of the federated learning model; the fifth level is used to indicate that federated learning model training is not supported, and data is not supported to be open to the outside world.
[0013] In a second aspect of the present disclosure, a federated learning server-side AI network element is provided, comprising: a memory; a processor, coupled to the memory, the processor being configured to execute instructions stored in the memory to implement a method as described in any of the above embodiments.
[0014] In a third aspect of the present disclosure, a communication system is provided, comprising: a federated learning server AI network element as described in any of the above embodiments; a registration network element, configured to discover candidate federated learning client AI network elements according to a federated learning client AI network element discovery request sent by the federated learning server AI network element, and send the federated learning client AI network element discovery result to the federated learning server AI network element; a federated learning client AI network element, configured to train the federated learning global model sent by the federated learning server AI network element to generate a local federated learning model, and send the local federated learning model to the federated learning server AI network element.
[0015] In some embodiments, the registration network element is configured to discover the candidate federated learning client AI network element corresponding to the data privacy level information based on the data privacy level information included in the discovery request.
[0016] In some embodiments, the federated learning client AI network element is configured to train the federated learning global model using a local data set.
[0017] In some embodiments, the federated learning client AI network element is configured to receive a shared data set sent by the federated learning server AI network element, and train the federated learning global model using the shared data set and the local data set.
[0018] In some embodiments, the federated learning client AI network element is configured to upload data associated with the data privacy level information of the federated learning client AI network element to the federated learning server AI network element.
[0019] In some embodiments, the registration network element is configured to process the registration of the federated learning member process based on a registration request sent by the federated learning member, wherein the federated learning member includes at least one of the federated learning server AI network element and the federated learning client AI network element, and the registration request includes data privacy level information of the federated learning member.
[0020] In some embodiments, the data privacy level of the federated learning members includes one of the following: the first level is used to indicate support for direct collection of arbitrary data; the second level is used to indicate support for local training and uploading of the federated learning model; the third level is used to indicate support for local training and uploading of the federated learning model, and support for uploading of part of the data that will contribute to the training of the federated learning model, wherein the privacy of the part of the data is lower than the specified level, and the amount of the part of the data is less than a predetermined proportion; the fourth level is used to indicate support for local training and uploading of the federated learning model; the fifth level is used to indicate support for federated learning model training and non-support for external opening of data.
[0021] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and when the instructions are executed by a processor, the method described in any of the above embodiments is implemented.
[0022] In a fifth aspect of the present disclosure, a computer program product is provided, comprising computer instructions, wherein when the computer instructions are executed by a processor, the method described in any of the above embodiments is implemented.
[0023] Other features and advantages of the present disclosure will become apparent from the following detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0025] Figure 1 A flowchart of a federated learning method according to an embodiment of the present disclosure;
[0026] Figure 2 This is a structural diagram of a federated learning server AI network element according to an embodiment of the present disclosure;
[0027] Figure 3 A schematic diagram of the structure of a communication system according to an embodiment of the present disclosure;
[0028] Figure 4 A flowchart of a federated learning method according to another embodiment of the present disclosure;
[0029] Figure 5 It is a schematic diagram of the structure of a communication system according to another embodiment of the present disclosure. DETAILED DESCRIPTION
[0030] The following will be combined with the 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 described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present disclosure and its application or use. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.
[0031] Unless specifically stated otherwise, the relative arrangement of components and steps, the numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present disclosure.
[0032] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.
[0033] Technologies, methods, and apparatus known to ordinary technicians in the relevant field may not be discussed in detail, but where appropriate, such technologies, methods, and apparatus should be considered part of the authorization specification.
[0034] In all examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.
[0035] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.
[0036] Figure 1 The following is a flow chart of a federated learning method according to an embodiment of the present disclosure. In some embodiments, the following federated learning method is performed by a federated learning server AI network element, including steps 11-16.
[0037] In step 11, a federated learning strategy is determined according to the business requirements included in the model training request sent by the user equipment.
[0038] In some embodiments, the federated learning strategy includes data privacy level information of federated learning members, and the proportion of each data privacy level in the data privacy level information.
[0039] It should be noted here that federated learning members include federated learning server AI network elements and federated learning client AI network elements.
[0040] For example, the data privacy level of AI network elements on the federated learning server side includes the following.
[0041] The first level, L1, is used to indicate that the AI network element on the federated learning server has no privacy and supports the direct collection of any data.
[0042] The second level, L2, is used to indicate that some data is private and supports local training and uploading of federated learning models.
[0043] For example, some data with low privacy (data volume higher than a certain percentage) can be uploaded. The data involved here may be original data or processed data, such as data obtained through operations such as knowledge distillation.
[0044] The third level, L3, is used to indicate that most data have a high degree of privacy, supports local training and uploading of federated learning models, and supports uploading of some data that will contribute to the training of federated learning models, where the privacy of some data is lower than the specified level and the amount of some data is less than the predetermined proportion.
[0045] The fourth level, L4, is used to indicate that local data privacy is strong and supports local training and uploading of federated learning models.
[0046] The fifth level, L5, indicates that local data has a high degree of privacy, does not support federated learning model training, and does not support data opening to the outside world.
[0047] For example, for requests with high performance requirements for the model, the federated learning server AI network element may decide to select clients with data privacy levels at level L2 or L3, sacrificing a certain degree of privacy to improve the performance of federated learning. For requests with high latency requirements, the federated learning server AI network element may decide to select clients at level L1 to directly collect data. For certain specific parameters, such as the area carried in the request, the server AI network element can only determine the specific training strategy based on the characteristics of the network elements in the relevant area.
[0048] In step 12, the registration network element is used to discover candidate federated learning client AI network elements.
[0049] In some embodiments, the step of discovering candidate federated learning client AI network elements using a registration network element includes the following.
[0050] 1) Send a federated learning client AI network element discovery request to the registered network element, where the discovery request includes data privacy level information, so that the registered network element can discover candidate federated learning client AI network elements corresponding to the data privacy level information.
[0051] 2) Receive the federated learning client AI network element discovery results sent by the registered network element.
[0052] In step 13, a federated learning client AI network element is selected from candidate federated learning client AI network elements according to the federated learning strategy.
[0053] It should be noted here that the members of each round of federated training may be inconsistent according to actual business needs. The federated learning strategy can be adjusted according to the actual training results of each round, such as adjusting the proportion of members at each data privacy level.
[0054] In step 14, the federated learning global model is sent to the federated learning client AI network element so that the federated learning client AI network element trains the federated learning global model to generate a local federated learning model.
[0055] In some embodiments, the federated learning client AI network element uses a local data set to train a federated learning global model to obtain a local federated learning model.
[0056] In some embodiments, a shared data set is sent to a federated learning client AI network element so that the federated learning client AI network element trains a federated learning global model using the shared data set and a local data set.
[0057] It should be noted here that the shared dataset integrates the data uploaded by members in the previous round of training. For traditional federated learning, only the model needs to be distributed, and there is no shared dataset.
[0058] In step 15, the local federated learning model uploaded by the federated learning client AI network element is aggregated.
[0059] In some embodiments, the federated learning client AI network element uploads data associated with its own data privacy level information to the federated learning server AI network element. In this case, the federated learning server AI network element integrates the data uploaded by the federated learning client AI network element.
[0060] For example, the data uploaded by the AI network element of the federated learning client can be raw data or processed data, such as data that has undergone knowledge distillation or encryption operations.
[0061] In step 16, the process of selecting a federated learning client AI network element is repeated until the federated learning is completed.
[0062] In some embodiments, the federated learning server AI network element sends a registration request to the registration network element, wherein the registration request includes data privacy level information of the federated learning server AI network element, so that the registration network element registers the federated learning server AI network element.
[0063] It should be noted that for AI network elements, data privacy level information is used to indicate data collection and federated learning related content. For other network elements that do not include AI capabilities, data privacy level only indicates operations related to data collection. In addition, for AI network elements, each AI scenario corresponds to a data privacy level.
[0064] In the federated learning method provided in the above embodiment of the present disclosure, the federated learning server-side AI network element selects a federated learning client AI network element from candidate federated learning client AI network elements for model training according to the federated learning strategy, thereby effectively improving the accuracy and generalization of the federated learning model and improving the user experience.
[0065] Figure 2 This is a structural diagram of a federated learning server AI network element according to an embodiment of the present disclosure.
[0066] like Figure 2 As shown, the federated learning server AI network element 20 can be in the form of a general-purpose computing device. The federated learning server AI network element 20 includes a memory 21, a processor 22, and a bus 23 connecting different system components.
[0067] The memory 21 may include, for example, a system memory, a non-volatile storage medium, etc. The system memory may store, for example, an operating system, an application, a boot loader, and other programs. The system memory may include a volatile storage medium, such as a random access memory (RAM) and / or a cache memory. The non-volatile storage medium may store, for example, instructions for a corresponding embodiment of at least one federated learning method being executed. The non-volatile storage medium includes, but is not limited to, a disk memory, an optical memory, a flash memory, and the like.
[0068] The processor 22 may be implemented by a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistors, and other discrete hardware components. Accordingly, each module such as an acquisition module, a calculation module, and an adjustment module may be implemented by a central processing unit (CPU) running instructions in a memory that execute corresponding steps, or may be implemented by a dedicated circuit that executes corresponding steps.
[0069] For example, the processor 22 is configured to execute instructions stored in the memory to implement the following Figure 1 The method of any one of the embodiments.
[0070] The bus 23 may use any of a variety of bus structures, including, but not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MCA) bus, and a Peripheral Component Interconnect (PCI) bus.
[0071] The interfaces 24, 25, 26 of the federated learning server AI network element 20, as well as the memory 21 and the processor 22 can be connected through a bus 23. The input and output interface 24 can provide a connection interface for input and output devices such as a display, a mouse, and a keyboard. The network interface 25 provides a connection interface for various networked devices. The storage interface 26 provides a connection interface for external storage devices such as a floppy disk, a USB flash drive, and an SD card.
[0072] Here, various aspects of the present disclosure are described with reference to flowcharts and / or block diagrams of methods, devices, and computer program products according to embodiments of the present disclosure. It should be understood that each frame of the flowchart and / or block diagram and the combination of frames can be implemented by computer-readable program instructions.
[0073] These computer-readable program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer or other programmable device to produce a machine, so that the processor executes the instructions to produce means for implementing the functions specified in one or more blocks in the flowchart and / or block diagram.
[0074] These computer-readable program instructions may also be stored in a computer-readable memory, which cause the computer to work in a specific manner to produce an article of manufacture, including instructions for implementing the functions specified in one or more blocks in the flowchart and / or block diagram.
[0075] The present disclosure can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects.
[0076] The present disclosure also provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and when the instructions are executed by a processor, the following is implemented: Figure 1 The method of any one of the embodiments.
[0077] The present disclosure also provides a computer program product, including computer instructions, wherein when the computer instructions are executed by a processor, the following is achieved: Figure 1 The method of any one of the embodiments.
[0078] Figure 3 The present invention is a schematic diagram of the structure of a communication system according to an embodiment of the present invention.
[0079] like Figure 3 As shown, the communication system includes a federated learning server AI network element 31, a registration network element 32, and a federated learning client AI network element 33. The federated learning server AI network element 31 is Figure 2 A federated learning server AI network element involved in any of the embodiments.
[0080] The registration network element 32 is configured to discover candidate federated learning client AI network elements according to the federated learning client AI network element discovery request sent by the federated learning server AI network element 31, and send the federated learning client AI network element discovery result to the federated learning server AI network element 31.
[0081] The federated learning client AI network element 33 is configured to train the federated learning global model issued by the federated learning server AI network element 31 to generate a local federated learning model, and send the local federated learning model to the federated learning server AI network element 31.
[0082] In some embodiments, the registration network element 32 is configured to discover candidate federated learning client AI network elements corresponding to the data privacy level information based on the data privacy level information included in the discovery request.
[0083] In some embodiments, the federated learning client AI network element 33 is configured to train the federated learning global model using a local data set.
[0084] In some embodiments, the federated learning client AI network element 33 is configured to receive a shared data set sent by the federated learning server AI network element 31, and use the shared data set and the local data set to train the federated learning global model.
[0085] In some embodiments, the federated learning client AI network element 33 is configured to upload data associated with its own data privacy level information to the federated learning server AI network element 31 .
[0086] In some embodiments, the registration network element 32 is configured to process the registration of the federated learning member process based on a registration request sent by the federated learning member, where the federated learning member includes at least one of a federated learning server AI network element and a federated learning client AI network element, and the registration request includes data privacy level information of the federated learning member.
[0087] In some embodiments, the data privacy level of federated learning members includes the following.
[0088] The first level, L1, is used to indicate that federated learning members have no privacy and supports the direct collection of arbitrary data.
[0089] The second level, L2, is used to indicate that some data is private and supports local training and uploading of federated learning models.
[0090] For example, some data with low privacy (data volume higher than a certain percentage) can be uploaded. The data involved here may be original data or processed data, such as data obtained through operations such as knowledge distillation.
[0091] The third level, L3, is used to indicate that most data have a high degree of privacy, supports local training and uploading of federated learning models, and supports uploading of some data that will contribute to the training of federated learning models, where the privacy of some data is lower than the specified level and the amount of some data is less than the predetermined proportion.
[0092] The fourth level, L4, is used to indicate that local data privacy is strong and supports local training and uploading of federated learning models.
[0093] The fifth level, L5, indicates that local data has a high degree of privacy, does not support federated learning model training, and does not support data opening to the outside world.
[0094] Figure 4 A flowchart of a federated learning method according to another embodiment of the present disclosure.
[0095] In step 40, the data privacy level of the network element is configured.
[0096] It should be noted here that the data privacy level of network elements includes the following.
[0097] The first level, L1, is used to indicate that the network element has no privacy and supports direct collection of arbitrary data.
[0098] The second level, L2, is used to indicate that some data is private and supports local training and uploading of federated learning models.
[0099] For example, some data with low privacy (data volume higher than a certain percentage) can be uploaded. The data involved here may be original data or processed data, such as data obtained through operations such as knowledge distillation.
[0100] The third level, L3, is used to indicate that most data have a high degree of privacy, supports local training and uploading of federated learning models, and supports uploading of some data that will contribute to the training of federated learning models, where the privacy of some data is lower than the specified level and the amount of some data is less than the predetermined proportion.
[0101] The fourth level, L4, is used to indicate that local data privacy is strong and supports local training and uploading of federated learning models.
[0102] The fifth level, L5, indicates that local data has a high degree of privacy, does not support federated learning model training, and does not support data opening to the outside world.
[0103] In step 41a, the federated learning server AI network element sends a registration request to the registration network element, where the registration request includes data privacy level information of the federated learning server AI network element so that the registration network element can register the federated learning server AI network element.
[0104] In step 41b, the federated learning client AI network element sends a registration request to the registration network element, where the registration request includes data privacy level information of the federated learning client AI network element so that the registration network element can register the federated learning client AI network element.
[0105] In step 41c, other network elements send a registration request to the registration network element, where the registration request includes data privacy level information of other network elements, so that the registration network element can register other network elements.
[0106] It should be noted here that for AI network elements, data privacy level information indicates data collection and federated learning related content. And for AI network elements, each AI scenario corresponds to a data privacy level. For other network elements, data privacy level information only indicates data collection related operations.
[0107] In step 42, the user equipment sends a model training request to the federated learning server AI network element.
[0108] In step 43, the federated learning server AI network element determines the federated learning strategy according to the business requirements included in the model training request.
[0109] For example, the federated learning strategy includes data privacy level information of federated learning members, and the proportion of each data privacy level in the data privacy level information.
[0110] For example, for requests with high performance requirements for the model, the federated learning server AI network element may decide to select clients with data privacy levels at level L2 or L3, sacrificing a certain degree of privacy to improve the performance of federated learning. For requests with high latency requirements, the federated learning server AI network element may decide to select clients at level L1 to directly collect data. For certain specific parameters, such as the area carried in the request, the server AI network element can only determine the specific training strategy based on the characteristics of the network elements in the relevant area.
[0111] In step 44, the federated learning server AI network element uses the registration network element to discover candidate federated learning client AI network elements.
[0112] For example, the federated learning server AI network element sends a federated learning client AI network element discovery request to the registration network element, wherein the discovery request includes data privacy level information, so that the registration network element discovers candidate federated learning client AI network elements corresponding to the data privacy level information. The registration network element sends the federated learning client AI network element discovery result to the federated learning server AI network element.
[0113] In step 45, the federated learning server AI network element selects a federated learning client AI network element from the candidate federated learning client AI network elements according to the federated learning strategy.
[0114] It should be noted here that the members of each round of federated training may be inconsistent according to actual business needs. The federated learning strategy can be adjusted according to the actual training results of each round, such as adjusting the proportion of members at each data privacy level.
[0115] In step 46, the federated learning server AI network element sends the federated learning global model to the federated learning client AI network element.
[0116] In step 47, the federated learning client AI network element trains the federated learning global model to generate a local federated learning model.
[0117] For example, the federated learning client AI network element uses the local data set to train the federated learning global model to obtain the local federated learning model.
[0118] For another example, the federated learning server AI network element sends a shared data set to the federated learning client AI network element, and the federated learning client AI network element uses the shared data set and the local data set to train the federated learning global model.
[0119] It should be noted here that the shared dataset integrates the data uploaded by members in the previous round of training. For traditional federated learning, only the model needs to be distributed, and there is no shared dataset.
[0120] In step 48, the federated learning client AI network element uploads the local federated learning model.
[0121] In step 49, the federated learning server AI network element aggregates the local federated learning model uploaded by the federated learning client AI network element.
[0122] In some embodiments, the federated learning client AI network element uploads data associated with its own data privacy level information to the federated learning server AI network element. In this case, the federated learning server AI network element integrates the data uploaded by the federated learning client AI network element.
[0123] For example, the data uploaded by the AI network element of the federated learning client can be raw data or processed data, such as data that has undergone knowledge distillation or encryption operations.
[0124] Repeat steps 45 to 49 until the federated learning is completed.
[0125] In step 410, the federated learning server AI network element sends a model training request response to the user equipment.
[0126] The present disclosure is described below through some specific embodiments. Embodiment 1:
[0127] In the scenario where multiple NWDAF (Network Data Analytics Function) network elements perform federated training in a 5G network, the server NWDAF and the client NWDAF carry their local data privacy level information when registering. The server NWDAF can use this information to flexibly and efficiently select federated learning members.
[0128] like Figure 5 As shown, the server NWDAF and the client NWDAF register with the NRF (Network Repository Function) network element. Each network element carries its data privacy level information when registering. This information indicates the privacy of the local data of the network element. The higher the data privacy level, the less data the network element can share.
[0129] When discovering client network elements, the server-side NWDAF obtains the data privacy level of each client network element and selects federated learning members based on actual business needs and the data privacy level of the client network element. When each round of federated training sends the model, the shared data set is sent based on the actual member's data privacy level and the shared data uploaded by the member. Embodiment 2:
[0130] In a 6G network, multiple AI network elements use a federated learning algorithm (such as the FEDFED algorithm) for federated training. Through the above embodiments of the present disclosure, the model performance and generalization of federated learning can be improved to meet user needs. Embodiment three:
[0131] In the scenario where 5G / 6G network AI network elements collect data, it is found that network elements with lower data privacy levels collect data.
[0132] By implementing the above-mentioned embodiments of the present disclosure, the following beneficial effects can be obtained.
[0133] 1) When a network element is registered, it carries the data privacy level information of the network element. With this information, the federated learning server can flexibly select federated learning members that can provide shared data according to business needs, or flexibly discover data sources without adding any complexity.
[0134] 2) 5G and 6G networks can support new federated learning algorithms and improve the performance and generalization of federated learning models.
[0135] In some embodiments, the functional unit described above may be implemented as a general-purpose processor, a programmable logic controller (PLC), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or any appropriate combination thereof, for performing the functions described in the present disclosure.
[0136] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.
[0137] The description of the present disclosure is given for the purpose of illustration and description, and is not intended to be exhaustive or to limit the present disclosure to the disclosed form. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present disclosure, and to enable those of ordinary skill in the art to understand the present disclosure and thereby design various embodiments with various modifications suitable for specific uses.
Claims
1. A federated learning method, executed by a federated learning server-side artificial intelligence AI network element, comprising: Determine a federated learning strategy based on the business requirements included in the model training request sent by the user device; Use the registered network elements to discover candidate federated learning client AI network elements; Selecting a federated learning client AI network element from among the candidate federated learning client AI network elements according to the federated learning strategy; Sending a federated learning global model to the federated learning client AI network element so that the federated learning client AI network element trains the federated learning global model to generate a local federated learning model; Aggregating the local federated learning model uploaded by the federated learning client AI network element; The process of selecting the federated learning client AI network element is repeated until the federated learning is completed.
2. The method according to claim 1, wherein: The federated learning strategy includes data privacy level information of federated learning members and the proportion of each data privacy level in the data privacy level information.
3. The method according to claim 2, wherein: The method of using the registration network element to discover the candidate federated learning client AI network element includes: Sending a federated learning client AI network element discovery request to the registration network element, wherein the discovery request includes the data privacy level information, so that the registration network element discovers the candidate federated learning client AI network element corresponding to the data privacy level information; Receive the federated learning client AI network element discovery result sent by the registration network element.
4. The method according to claim 1, wherein: The local federated learning model is obtained by the federated learning client AI network element training the federated learning global model using a local data set.
5. The method according to claim 1, further comprising: A shared data set is issued to the federated learning client AI network element so that the federated learning client AI network element trains the federated learning global model using the shared data set and the local data set.
6. The method according to claim 1, further comprising: Integrate the data uploaded by the federated learning client AI network element.
7. The method according to any one of claims 1 to 6, further comprising: Send a registration request to the registration network element, wherein the registration request includes data privacy level information of the federated learning server AI network element.
8. The method according to claim 7, wherein: The data privacy level of the federated learning server AI network element includes one of the following: the first level is used to indicate support for direct collection of any data; the second level is used to indicate support for local training and uploading of the federated learning model; The third level is used to indicate support for local training and uploading of federated learning models, and support for uploading of some data that will be helpful for training the federated learning models, where the privacy of the said part of data is lower than the specified level, and the amount of the said part of data is less than the predetermined proportion; the fourth level is used to indicate support for local training and uploading of federated learning models; The fifth level is used to indicate that federated learning model training is not supported and data is not open to the outside world.
9. A federated learning server AI network element, comprising: Memory; A processor is coupled to the memory, and the processor is configured to execute the method according to any one of claims 1 to 8 based on instructions stored in the memory.
10. A communication system comprising: Comprising the federated learning server AI network element as claimed in claim 9; A registration network element, configured to discover candidate federated learning client AI network elements according to the federated learning client AI network element discovery request sent by the federated learning server AI network element, and send the federated learning client AI network element discovery result to the federated learning server AI network element; The federated learning client AI network element is configured to train the federated learning global model issued by the federated learning server AI network element to generate a local federated learning model, and send the local federated learning model to the federated learning server AI network element.
11. The system according to claim 10, wherein: The registration network element is configured to discover the candidate federated learning client AI network element corresponding to the data privacy level information according to the data privacy level information included in the discovery request.
12. The system according to claim 10, wherein: The federated learning client AI network element is configured to train the federated learning global model using a local data set.
13. The system according to claim 10, wherein: The federated learning client AI network element is configured to receive a shared data set sent by the federated learning server AI network element, and train the federated learning global model using the shared data set and a local data set.
14. The system according to claim 10, wherein: The federated learning client AI network element is configured to upload data associated with the data privacy level information of the federated learning client AI network element to the federated learning server AI network element.
15. The system according to any one of claims 10 to 14, wherein: The registration network element is configured to process the registration of the federated learning member process according to a registration request sent by the federated learning member, wherein the federated learning member includes at least one of the federated learning server AI network element and the federated learning client AI network element, and the registration request includes data privacy level information of the federated learning member.
16. The method according to claim 15, wherein: The data privacy level of the federated learning member includes one of the following: the first level is used to indicate support for direct collection of any data; the second level is used to indicate support for local training and uploading of the federated learning model; The third level is used to indicate support for local training and uploading of federated learning models, and support for uploading of some data that will be helpful for training the federated learning models, where the privacy of the said part of data is lower than the specified level, and the amount of the said part of data is less than the predetermined proportion; the fourth level is used to indicate support for local training and uploading of federated learning models; The fifth level is used to indicate that federated learning model training is not supported and data is not open to the outside world.
17. A computer-readable storage medium, wherein: The computer-readable storage medium stores computer instructions, and when the instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.
18. A computer program product, comprising computer instructions, wherein when the computer instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.