Method, device, electronic device and readable storage medium for constructing a federated learning system

Abstract

The embodiments of the present application provide a method, device, electronic device and readable storage medium for constructing a federated learning system, and relate to the field of computer technology. The method includes: obtaining federated learning task construction information, and determining the associated candidate participating devices based on the federated learning task construction information, and then sending a federated learning request containing the federated learning task construction information to each candidate participating device. Reply information returned by each candidate participating device can be received, and the target participating device can be determined based on the reply information, and then a federated learning task can be constructed based on the federated learning task construction information, and the federated learning task can be performed with at least one target participating device. The present application can flexibly and specifically determine participating devices according to business needs through the federated learning task construction information, so that a larger number of participating devices are determined, and the data provided by each participating device is more in line with the training requirements of the federated learning task, and finally the training results of the federated learning task are optimized.

Description

Technical Field

[0001] The present application relates to the field of computer technology. Specifically, the present application relates to a method, device, electronic device and readable storage medium for constructing a Bang learning system. Background Art

[0002] Federated learning is a machine learning system that has been gradually applied in the fields of finance, medical care, transportation, communications, etc. It can protect user privacy and strengthen data security. Through distributed joint model training, multiple participants can provide data separately on the basis of ensuring data privacy security and legal compliance, and achieve joint modeling, thereby solving the problem of data silos.

[0003] Typically, the participants of a federated learning task are set or configured in advance, or are determined based on historical feedback data from participants in federated learning. For example, the participants for the next execution of a federated learning task are determined by obtaining feedback data from multiple participants after they have historically accepted federated learning invitations.

[0004] When the business targeted by the federated learning task changes, the required business data will also change. At this time, different participants are needed to provide new business data. However, in the current technical solution, the method of determining participants is not flexible and accurate enough. If the participants in the historical tasks are used or screened from the participants in the historical tasks, the determined participants may not match the changed federated learning tasks, making it difficult to meet actual application needs. Summary of the invention

[0005] The purpose of the embodiments of the present application is to solve the problem that the method for determining federated learning participants is not flexible and accurate enough.

[0006] In a first aspect, the present application provides a method for constructing a federated learning system, the method comprising:

[0007] Acquire federated learning task construction information, and determine at least one associated candidate participating device based on the federated learning task construction information;

[0008] Send a federated learning request containing federated learning task construction information to each candidate participating device;

[0009] receiving reply information returned by at least one candidate participating device, and determining at least one target participating device from the at least one candidate participating device based on the reply information;

[0010] A federated learning task is constructed based on the federated learning task construction information, and the federated learning task is executed with at least one target participating device.

[0011] In an optional embodiment of the first aspect, determining at least one associated candidate participating device based on the federated learning task construction information includes:

[0012] Determine the corresponding federated learning type based on the federated learning task construction information;

[0013] At least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information; the federated learning type is used to describe that the business data characteristics or user data characteristics of at least one candidate participating device are similar.

[0014] In an optional embodiment of the first aspect, determining a corresponding federated learning type based on the federated learning task construction information includes:

[0015] Extracting the federated learning task type from the federated learning task construction information; there is a corresponding relationship between the federated learning task type and the federated learning type;

[0016] The federated learning type corresponding to the federated learning task construction information is determined based on the corresponding relationship and the federated learning task type.

[0017] In an optional embodiment of the first aspect, the federated learning type includes horizontal federated learning;

[0018] When the federated learning type is horizontal federated learning, at least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information, including:

[0019] Extract business data features from the federated learning task construction information, and use devices with business data features as candidate participating devices.

[0020] In an optional embodiment of the first aspect, the federated learning type includes vertical federated learning;

[0021] When the federated learning type is vertical federated learning, at least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information, including:

[0022] Extract user data features from the federated learning task construction information, and use devices corresponding to the user data features as candidate participating devices.

[0023] In an optional embodiment of the first aspect, the method further comprises:

[0024] Sending federated learning application information to the federated learning coordination device, where the federated learning application information includes federated learning task construction information;

[0025] Receive the federated learning request sent by the federated learning coordination device, and return the reply information to the federated learning coordination device; if the reply information is confirmed, it will perform the federated learning task as one of the target participating devices together with the federated learning coordination device and other target participating devices.

[0026] In an optional embodiment of the first aspect, the federated learning request also includes at least one of the following: resource information, computing power information, and energy consumption information required to participate in the federated learning.

[0027] In an optional embodiment of the first aspect, when each candidate participating device receives a federated learning request, the following operations are performed:

[0028] Determine the degree of relevance to the federated learning request based on at least one of the required resource information, computing power information, and energy consumption information in the federated learning request, and the federated learning task construction information;

[0029] If the relevance is greater than or equal to the preset threshold, the response information returned to the federated learning request is confirmation;

[0030] If the relevance is less than the preset threshold, the response information returned to the federated learning request is rejection.

[0031] In a second aspect, the present application provides a method for constructing a federated learning system, which is applied to a federated learning coordination device, and the method includes:

[0032] Receive federated learning application information, which includes federated learning task construction information;

[0033] Determine at least one associated candidate participating device based on the federated learning task construction information, where the at least one candidate participating device includes a device that sends the federated learning application information;

[0034] Send a federated learning request containing federated learning task construction information to each candidate participating device;

[0035] receiving reply information returned by at least one candidate participating device, and determining at least one target participating device from the at least one candidate participating device based on the reply information;

[0036] A federated learning task is constructed based on the federated learning task construction information, and the federated learning task is executed together with at least one target participating device.

[0037] In an optional embodiment of the second aspect, determining at least one associated candidate participating device based on the federated learning task construction information includes:

[0038] Determine the corresponding federated learning type based on the federated learning task construction information;

[0039] At least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information; the federated learning type is used to describe that the business data characteristics or user data characteristics of at least one candidate participating device are similar.

[0040] In an optional embodiment of the second aspect, determining a corresponding federated learning type based on the federated learning task construction information includes:

[0041] Extracting the federated learning task type from the federated learning task construction information; there is a corresponding relationship between the federated learning task type and the federated learning type;

[0042] The federated learning type corresponding to the federated learning task construction information is determined based on the corresponding relationship and the federated learning task type.

[0043] In an optional embodiment of the second aspect, the federated learning type includes horizontal federated learning;

[0044] When the federated learning type is horizontal federated learning, at least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information, including:

[0045] Extract business data features from the federated learning task construction information, and use devices with business data features as candidate participating devices.

[0046] In an optional embodiment of the second aspect, the federated learning type includes vertical federated learning;

[0047] When the federated learning type is vertical federated learning, at least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information, including:

[0048] Extract user data features from the federated learning task construction information, and use devices corresponding to the user data features as candidate participating devices.

[0049] In a third aspect, a federated learning system construction device is provided, the device comprising:

[0050] A candidate device determination module, configured to obtain federated learning task construction information and determine at least one associated candidate participating device based on the federated learning task construction information;

[0051] A learning request sending module is used to send a federated learning request containing federated learning task construction information to each candidate participating device;

[0052] A target device determining module, configured to receive reply information returned by at least one candidate participating device, and determine at least one target participating device from the at least one candidate participating device based on the reply information;

[0053] The federated learning module is used to construct a federated learning task based on the federated learning task construction information, and execute the federated learning task with at least one target participating device.

[0054] In an optional embodiment of the third aspect, when the candidate device determination module determines at least one associated candidate participating device based on the federated learning task construction information, it is specifically configured to:

[0055] Determine the corresponding federated learning type based on the federated learning task construction information;

[0056] At least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information; the federated learning type is used to describe that the business data characteristics or user data characteristics of at least one candidate participating device are similar.

[0057] In an optional embodiment of the third aspect, when the candidate device determination module determines the corresponding federated learning type based on the federated learning task construction information, it is specifically used to:

[0058] Extracting the federated learning task type from the federated learning task construction information; there is a corresponding relationship between the federated learning task type and the federated learning type;

[0059] The federated learning type corresponding to the federated learning task construction information is determined based on the corresponding relationship and the federated learning task type.

[0060] In an optional embodiment of the third aspect, the federated learning type includes horizontal federated learning;

[0061] When the federated learning type is horizontal federated learning, the candidate device determination module is specifically used to determine at least one associated candidate participating device based on the federated learning type and the federated learning task construction information:

[0062] Extract business data features from the federated learning task construction information, and use devices with business data features as candidate participating devices.

[0063] In an optional embodiment of the third aspect, the federated learning type includes vertical federated learning;

[0064] When the federated learning type is vertical federated learning, the candidate device determination module is specifically used to determine at least one associated candidate participating device based on the federated learning type and the federated learning task construction information:

[0065] Extract user data features from the federated learning task construction information, and use devices corresponding to the user data features as candidate participating devices.

[0066] In an optional embodiment of the third aspect, an application information sending module is further included, which is specifically used to:

[0067] Sending federated learning application information to the federated learning coordination device, where the federated learning application information includes federated learning task construction information;

[0068] Receive the federated learning request sent by the federated learning coordination device, and return the reply information to the federated learning coordination device; if the reply information is confirmed, it will perform the federated learning task as one of the target participating devices together with the federated learning coordination device and other target participating devices.

[0069] In an optional embodiment of the third aspect, the federated learning request also includes at least one of the following: resource information, computing power information, and energy consumption information required to participate in federated learning.

[0070] In an optional embodiment of the third aspect, when each candidate participating device receives a federated learning request, the following operations are performed:

[0071] Determine the degree of relevance to the federated learning request based on at least one of the required resource information, computing power information, and energy consumption information in the federated learning request, and the federated learning task construction information;

[0072] If the relevance is greater than or equal to the preset threshold, the response information returned to the federated learning request is confirmation;

[0073] If the relevance is less than the preset threshold, the response information returned to the federated learning request is rejection.

[0074] In a fourth aspect, a federated learning system construction device is provided, the device comprising:

[0075] A first device determination module is configured to receive federated learning application information, wherein the federated learning application information includes federated learning task construction information; determine at least one associated candidate participating device based on the federated learning task construction information, wherein the at least one candidate participating device includes a device that sends the federated learning application information;

[0076] A first request sending module, used to send a federated learning request containing federated learning task construction information to each candidate participating device;

[0077] A second device determination module, configured to receive reply information returned by at least one candidate participating device, and determine at least one target participating device from the at least one candidate participating device based on the reply information;

[0078] The first federated learning module is used to construct a federated learning task based on the federated learning task construction information, and execute the federated learning task with at least one target participating device.

[0079] In a fifth aspect, an electronic device is provided, the electronic device comprising:

[0080] A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for constructing a federated learning system of any of the above embodiments is implemented.

[0081] In a sixth aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method for constructing a federated learning system of any of the above embodiments is implemented.

[0082] The above-mentioned method for constructing a federated learning system obtains federated learning task construction information, and determines at least one associated candidate participating device based on the federated learning task construction information, and then sends a federated learning request containing the federated learning task construction information to each candidate participating device. After receiving the reply information returned by at least one candidate participating device, at least one target participating device can be determined from at least one candidate participating device based on the reply information, and then a federated learning task is constructed based on the federated learning task construction information, and the federated learning task is performed with at least one target participating device. The present application can flexibly and specifically determine participating devices according to business needs through the federated learning task construction information, thereby achieving a larger number of determined participating devices, and the data provided by each participating device is more in line with the training requirements of the federated learning task, and ultimately optimizing the training results of the federated learning task. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in describing the embodiments of the present application.

[0084] Figure 1 A flowchart of a method for constructing a federated learning system provided in an embodiment of the present application;

[0085] Figure 2 A schematic diagram of the interaction between various devices in a method for building a federated learning system provided in an embodiment of the present application;

[0086] Figure 3 A schematic diagram of the interaction between various devices in a method for building a federated learning system provided in an embodiment of the present application;

[0087] Figure 4 A schematic diagram of the interaction between various devices in a method for building a federated learning system provided in an embodiment of the present application;

[0088] Figure 5 A schematic diagram of the interaction between various devices in a method for building a federated learning system provided in an embodiment of the present application;

[0089] Figure 6A schematic diagram of the structure of a federated learning system construction device provided in an embodiment of the present application;

[0090] Figure 7 A schematic diagram of the structure of a federated learning system construction device provided in an embodiment of the present application;

[0091] Figure 8 A schematic diagram of the structure of an electronic device constructed for a federated learning system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0092] The embodiments of the present application are described below in conjunction with the drawings in the present application. It should be understood that the implementation methods described below in conjunction with the drawings are exemplary descriptions for explaining the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions of the embodiments of the present application.

[0093] It will be understood by those skilled in the art that, unless specifically stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the terms "including" and "comprising" used in the embodiments of the present application refer to that the corresponding features can be implemented as the presented features, information, data, steps, operations, elements and / or components, but do not exclude the implementation as other features, information, data, steps, operations, elements, components and / or combinations thereof supported by the technical field. It should be understood that when we say that an element is "connected" or "coupled" to another element, the one element may be directly connected or coupled to the other element, or it may refer to that the one element and the other element establish a connection relationship through an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein indicates at least one of the items defined by the term, for example, "A and / or B" may be implemented as "A", or as "B", or as "A and B".

[0094] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0095] Federated learning is essentially a distributed machine learning technology, or a machine learning framework. Its goal is to achieve joint modeling and improve the effectiveness of artificial intelligence models while ensuring data privacy, security, legality and compliance.

[0096] Federated learning can protect user privacy and strengthen data security by allowing multiple participants to provide data separately through distributed joint model training, breaking down data barriers and thus solving the problem of data silos. In current federated learning solutions, participants in federated learning are generally set or configured in advance, or determined based on historical data participating in federated learning. They cannot be well adapted to the needs of various business changes, and it is difficult to flexibly determine more and more suitable participants in a more targeted manner according to business needs.

[0097] The federated learning system construction method, device, electronic device and readable storage medium provided in this application are intended to solve the above technical problems of the prior art.

[0098] The following describes several exemplary embodiments to illustrate the technical solutions of the embodiments of the present application and the technical effects produced by the technical solutions of the present application. It should be noted that the following embodiments can refer to, draw on or combine with each other, and the same terms, similar features and similar implementation steps in different embodiments will not be described repeatedly.

[0099] The federated learning system construction provided in the embodiments of the present application can be applied to terminal devices.

[0100] Technicians in this technical field can understand that the "terminal device" used here can be a mobile phone, a tablet computer, a PDA (Personal Digital Assistant), a MID (Mobile Internet Device), etc.; the "server" can be implemented by an independent server or a server cluster consisting of multiple servers.

[0101] The present application provides a method for constructing a federated learning system. Figure 1 As shown, it can be applied to a federated learning initiator device, and the method may include:

[0102] Step S101: obtain federated learning task construction information, and determine at least one associated candidate participating device based on the federated learning task construction information.

[0103] In an embodiment of the present application, the federated learning system may include three types of devices: an initiating device, a coordinating device, and a participating device. The initiating device refers to a device that initiates federated learning, the participating device may be a device that provides data for federated learning (when the initiating device also participates in federated learning, the participating device may include the initiating device), and the coordinator is a device that manages all participating devices and performs operations such as model aggregation and distribution and model interaction.

[0104] Among them, the initiating device, the coordinating device and the participating device can be a terminal device, a server, a base station, a network management device, a core network element, etc.

[0105] In some embodiments, the initiating device, the coordinating device, and the participating devices may also be referred to as the initiator, the coordinator, and the participant in the federated learning system.

[0106] Federated learning systems can be roughly divided into federated learning systems with coordination devices and federated learning systems without coordination devices.

[0107] When the federated learning system is a federated learning system without a coordination device, the terminal device can serve as an initiator of the federated learning system, obtain federated learning task construction information, and determine at least one associated candidate participating device based on the federated learning task construction information.

[0108] In an embodiment of the present application, determining at least one associated candidate participating device based on the federated learning task construction information may include the following steps: determining the corresponding federated learning type based on the federated learning task construction information; determining at least one associated candidate participating device based on the federated learning type and the federated learning task construction information; the federated learning type is used to describe the similarity of business data characteristics or user data characteristics of at least one candidate participating device.

[0109] The federated learning task construction information may be any relevant information describing the federated learning task to be constructed. Specifically, the federated learning task construction information may include one or more of the federated learning task type, business data characteristics, and user data characteristics.

[0110] For each federated learning task to be constructed, the federated learning task type can be used to describe the purpose, subject, corresponding business or federated learning object of the federated learning task. Federated learning tasks can be classified by various federated learning task types. In a specific application scenario, multiple federated learning task types can be pre-set, and then the federated learning task type corresponding to the federated learning task to be constructed can be selected from the multiple federated learning task types.

[0111] In the embodiment of the present application, determining the corresponding federated learning type based on the federated learning task construction information may include the following steps:

[0112] Extracting the federated learning task type from the federated learning task construction information; there is a corresponding relationship between the federated learning task type and the federated learning type;

[0113] The federated learning type corresponding to the federated learning task construction information is determined based on the corresponding relationship and the federated learning task type.

[0114] The federated learning type may be horizontal federated learning or vertical federated learning. The federated learning task type corresponding to each federated learning type may be pre-set, and based on the corresponding relationship, the federated learning type corresponding to the federated learning task to be constructed may be determined based on the federated learning task type in the federated learning task construction information.

[0115] For example, the federated learning task type may be a user terminal experience test task. In the preset correspondence, the user terminal experience test task corresponds to vertical federated learning, and the federated learning type corresponding to the federated learning task to be constructed is vertical federated learning.

[0116] In some embodiments, the type of federated learning task corresponding to the federated learning task to be constructed can also be determined by the content in the federated learning task construction information. For example, if the federated learning task construction information includes user data features (such as data source information, user information, etc.), the type of federated learning task corresponding to the federated learning task construction information is vertical federated learning; if the federated learning task construction information includes business data features, the type of federated learning task corresponding to the federated learning task construction information is horizontal federated learning; if both are included, the corresponding federated learning type can be further determined based on the federated learning task type.

[0117] In an embodiment of the present application, when the federated learning type is horizontal federated learning, determining at least one associated candidate participating device based on the federated learning type and the federated learning task construction information may include: extracting business data features in the federated learning task construction information, and selecting devices with business data features as candidate participating devices.

[0118] Among them, business data features are used to describe the features of the business data targeted by the federated learning task to be constructed. When the federated learning type is horizontal federated learning, it means that we need to screen out devices with similar business data features as candidate participating devices. The data provided by these candidate participating devices can better support this federated learning.

[0119] For example, when a terminal device as an initiating device takes an image (or video) of a scenic spot and needs to use federated learning technology to improve the clarity of the photo or video, the federated learning task type in the federated learning task construction information is an image processing task (or video processing task), and the business data feature is the image feature (or video feature) extracted from the scenic spot image (or video). Based on the federated learning task type being an image processing task (or video processing task), it is determined that the type of this federated learning task is horizontal federated learning, and at least one associated candidate participating device is determined based on the extracted image feature (or video feature). Specifically, the device having the image feature (or video feature) can be used as a candidate participating device.

[0120] The extracted image features include but are not limited to geographic information, weather information, and character information. Based on the extracted image features, a device having image data with the image features can be determined as a candidate participating device. The method of extracting image features can be an artificial intelligence method, which is not limited in this application.

[0121] In this example, the determined participating device may be a terminal device that performs a photographing operation on the scenic spot corresponding to the geographic information in the image feature, a public cloud device that stores photos of the scenic spot, and one or more of other related applications.

[0122] In an embodiment of the present application, when the federated learning type is vertical federated learning, determining at least one associated candidate participating device based on the federated learning type and the federated learning task construction information may include: extracting user data features in the federated learning task construction information, and taking the device corresponding to the user data features as a candidate participating device.

[0123] The user data feature may be at least one of data source information and user information. The data source information may refer to device attribute information of one or more groups of devices that meet preset conditions. For example, the data source information may be a group of device identifiers corresponding to a group of user terminal devices, and the user terminal devices corresponding to the group of device identifiers may be used as candidate participating devices.

[0124] The data source information may also be location information (province, city or street, etc.) or device model information, and the devices corresponding to the location information and device model information may be selected as candidate participating devices.

[0125] For example, when the base station is used as the initiating device, the federated learning task type in the federated learning task construction information can be a user device experience analysis task, and the determined federated learning type is vertical federated learning. The user data features in the federated learning task construction information can be the source information of the data in the base station (or the device information served by the base station) or the specified user device ID. At least one possible participating device can be determined based on the data source information or the user device ID. Specifically, the device corresponding to the data source information or the user device ID can be used as a candidate participating device.

[0126] Step S102: Send a federated learning request including federated learning task construction information to each candidate participating device.

[0127] Specifically, the federated learning request can be a NAS (Non-Access Stratum) message, an F1-AP message, an RRC (third layer information that processes the control plane between UE (User Equipment) and eNodeB (Evolved Node-B)) message, a management message, or an application message, etc. Different forms of federated learning requests can be sent for different types of candidate participating devices. For example, the core network may send a NAS message to the user equipment (ie, UE), the base station may send an F1-AP message to the core network, the base station may send an RRC message to the user equipment, the network management module may send a management message, and the application module or the application module may send an application message, etc.

[0128] In an embodiment of the present application, the federated learning request also includes at least one of the following: resource information, computing power information, and energy consumption information required to participate in federated learning.

[0129] In the embodiment of the present application, when each candidate participating device receives the federated learning request, it performs the following operations:

[0130] Determine the degree of relevance to the federated learning request according to the required resource information in the federated learning request, at least one of the computing power information and the energy consumption information, and the federated learning task construction information;

[0131] If the relevance is greater than or equal to the preset threshold, the response information returned to the federated learning request is confirmation;

[0132] If the relevance is less than the preset threshold, the response information returned to the federated learning request is rejection.

[0133] Among them, each candidate participating device can determine the degree of relevance to the federated learning request based on the required resource information in the federated learning request, at least one of the computing power information and energy consumption information, and the federated learning task construction information. If the degree of relevance is greater than or equal to a preset threshold, the federated learning request is accepted; if the degree of relevance is less than the preset threshold, the federated learning request is rejected.

[0134] Specifically, the method for determining the degree of relevance to the federated learning request may include at least one of the following:

[0135] (1) Match the federated learning task type, business data features or user data features in the federated learning task construction information with the data in the device and calculate the degree of correlation. For example, when the business data feature is an image feature, match the image feature with the image data of the candidate participating device. For example, calculate the similarity between the image feature of the extracted image data of the candidate parameter device and the image feature in the federated learning task request as the degree of correlation. If the similarity is greater than or equal to a preset threshold, then the degree of correlation is considered to be greater than or equal to the preset threshold;

[0136] (2) Correlation between calculation parameters: Each candidate participating device compares the resource utilization, computing power parameters and energy consumption parameters that it can provide with the resource information, computing power information and energy consumption information in the federated learning request. If at least two of the resource utilization, computing power parameters and energy consumption parameters that it can provide can meet the requirements in the federated learning request, it can be determined that the correlation with the federated learning request is greater than or equal to a preset threshold.

[0137] As an optional implementation method, for the resource information, computing power information and energy consumption information in the federated learning request, the candidate participating device can set corresponding thresholds for each item respectively. For example, the threshold for resource information (such as resource utilization) can be set as the first threshold, the threshold for computing power information can be set as the second threshold, and the threshold for energy consumption information can be set as the third threshold. If any one of the resource information, computing power information and energy consumption information is greater than the corresponding threshold, the federated learning request is rejected, otherwise the request is accepted. When comparing the thresholds, the current usage status of the device's resources, computing power and energy consumption can also be considered. For example, the candidate participating device can add the currently used resource utilization to the resource utilization in the federated learning request, and then compare it with the first threshold. If it is greater than, the federated learning request is rejected, otherwise, the request is accepted.

[0138] The above implementation method is only an example, and this application does not limit the method of determining the degree of relevance to the federated learning request.

[0139] Step S103: receiving reply information returned by at least one candidate participating device, and determining at least one target participating device from the at least one candidate participating device based on the reply information.

[0140] Specifically, the candidate participating device whose reply information is confirmation may be determined as the target participating device.

[0141] Step S104: construct a federated learning task based on the federated learning task construction information, and execute the federated learning task with at least one target participating device.

[0142] The above-mentioned method for constructing a federated learning system obtains federated learning task construction information, and determines at least one associated candidate participating device based on the federated learning task construction information, and then sends a federated learning request containing the federated learning task construction information to each candidate participating device. After receiving the reply information returned by at least one candidate participating device, at least one target participating device can be determined from at least one candidate participating device based on the reply information, and then a federated learning task is constructed based on the federated learning task construction information, and the federated learning task is performed with at least one target participating device. The present application can flexibly and specifically determine participating devices according to business needs through the federated learning task construction information, thereby achieving a larger number of determined participating devices, and the data provided by each participating device is more in line with the training requirements of the federated learning task, and ultimately optimizing the training results of the federated learning task.

[0143] A possible implementation method is provided in an embodiment of the present application. When a coordination device exists, the method for constructing a federated learning system provided by the present application further includes:

[0144] (1) The device serving as the initiating device sends federated learning application information to the federated learning coordination device, where the federated learning application information includes federated learning task construction information;

[0145] (2) Receive a federated learning request sent by a federated learning coordination device, and return a reply message to the federated learning coordination device in response to the federated learning request; if the reply message is a confirmation, the device acts as one of the target participating devices and performs the federated learning task together with the federated learning coordination device and other target participating devices.

[0146] In the embodiment of the present application, the federated learning coordination device can perform operations such as determining participating devices, model aggregation and distribution, and model interaction.

[0147] Specifically, the federated learning coordination device can be used to perform the following operations:

[0148] Receive federated learning application information, which includes federated learning task construction information;

[0149] Determine at least one associated candidate participating device based on the federated learning task construction information, where the at least one candidate participating device includes a device that sends the federated learning application information;

[0150] Send a federated learning request containing federated learning task construction information to each candidate participating device;

[0151] receiving reply information returned by at least one candidate participating device, and determining at least one target participating device from the at least one candidate participating device based on the reply information;

[0152] A federated learning task is constructed based on the federated learning task construction information, and the federated learning task is executed together with at least one target participating device.

[0153] In an embodiment of the present application, determining at least one associated candidate participating device based on the federated learning task construction information includes:

[0154] Determine the corresponding federated learning type based on the federated learning task construction information;

[0155] At least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information; the federated learning type is used to describe that the business data characteristics or user information of at least one candidate participating device are similar.

[0156] Among them, the corresponding federated learning type is determined based on the federated learning task construction information, including:

[0157] Extracting the federated learning task type from the federated learning task construction information; there is a corresponding relationship between the federated learning task type and the federated learning type;

[0158] The federated learning type corresponding to the federated learning task construction information is determined based on the corresponding relationship and the federated learning task type.

[0159] In the embodiment of the present application, when the federated learning type is horizontal federated learning, determining at least one associated candidate participating device based on the federated learning type and the federated learning task construction information includes:

[0160] Extract business data features from the federated learning task construction information, and use devices with business data features as candidate participating devices.

[0161] In some other embodiments, when the federated learning type is vertical federated learning, determining at least one associated candidate participating device based on the federated learning type and the federated learning task construction information includes:

[0162] Extract user data features from the federated learning task construction information, and use devices corresponding to the user data features as candidate participating devices.

[0163] The specific implementation method for determining the candidate participating devices is consistent with the technical solutions in step S101 to step S104 and will not be described again here.

[0164] The following is a specific example to illustrate the method for constructing a federated learning system provided by the present application. In the following examples, the initiating device, the coordinating device, and the participating device are referred to as the initiator, the coordinator, and the participating device in the federated learning system.

[0165] In actual application scenarios, federated learning systems can be roughly divided into federated learning systems with coordinators and federated learning systems without coordinators.

[0166] Among them, the federated learning tasks in each federated learning system can be divided into two types: horizontal federated learning and vertical federated learning, which will be explained one by one later.

[0167] Example 1: Type 1: When the federated learning system is a situation without a coordinator, and the federated learning type corresponding to the federated learning task to be constructed is vertical federated learning, the device interaction diagram of the method for constructing a federated system provided in this application can be as follows: Figure 2 shown.

[0168] In step 1, base station 1 may serve as an initiator, obtain federated learning task construction information, and determine at least one associated candidate participating device based on the federated learning task construction information.

[0169] Specifically, the federated learning task construction information may include the federated learning task type, data source information, and user information (such as user ID). The federated learning task type may be an analysis and prediction task for user device experience.

[0170] Base station 1 may determine at least one associated candidate participating device based on the type of federated learning task, and one or more of the data source information and the user information. The determined candidate participating devices may include a user equipment (UE) group corresponding to the data stored in base station 1, a base station having data of the group of UEs or serving the group of UEs (e.g., base station 1 and base station 2), and a core network element, etc.

[0171] In step 2, base station 1 may send a federated learning request message (i.e., a federated learning request) to the candidate participating device. The request message may be a NAS message, an F1-AP message, or an RRC message. The request message may include one or more of the federated learning task type, data source information, and user information, and may also include information such as resources, computing power, and energy consumption required to participate in the federated learning service.

[0172] Step 3: After receiving the request message for establishing federated learning, each candidate participating device may determine to accept or reject the request message.

[0173] Specifically, each candidate participating device may agree or reject the federated learning request message according to the content of the request message. For example, the candidate participating device may determine the relevance between the federated learning task to be constructed and itself according to one or more of the federated learning task type and business data characteristics. If the relevance is less than a preset value, the federated learning request message is rejected, otherwise the request message is accepted.

[0174] In some embodiments, the candidate participating device can determine whether it meets the resources, computing power, energy consumption, etc. required by the federated learning task in the message request based on its current resources, computing power, energy consumption and other parameters. If so, it accepts the request message, otherwise it rejects it.

[0175] In other embodiments, a corresponding single threshold value may also be set for each parameter in the resource, computing power, and energy consumption information required for the federated learning task in the message request. For example, when the resource utilization is greater than the preset resource utilization threshold, the request message is rejected, otherwise the request message is accepted. For another example, the candidate participating device can compare the total resource utilization of the currently used resources plus the resources required for the federated learning service with the resource utilization it can provide. If it is greater than the resource utilization it can provide, the request message is rejected, otherwise the request message is accepted.

[0176] In some embodiments, the user corresponding to the user device or application has the right to decide whether to participate in the federated learning task. If the user does not want to participate in the federated learning task, the user can also reject the request message through control operations.

[0177] Step 4: Each candidate participating device sends a confirmation message to base station 1, which confirms that the candidate participating device accepts the request as the target participating device that ultimately participates in federated learning. Base station 1 can construct a federated learning task based on the federated learning task construction information and perform the federated learning task with each target participating device.

[0178] like Figure 2 As shown, the confirmation message returned by the user device group and the core network element is to accept the request, and the confirmation message returned by base station 2 is to reject the request. At this time, the target participating devices determined are the user device group and the core network element. Base station 1 can construct a federated learning task based on the federated learning task construction information, and the user device group and the core network element jointly participate in the execution of the federated learning task.

[0179] Example 2: Type 2: When the federated learning system is a situation without a coordinator, and the federated learning type corresponding to the federated learning task to be constructed is horizontal federated learning, the device interaction diagram of the method for constructing a federated system provided in this application can be as follows: Figure 3 shown.

[0180] In step 1, user equipment 1 (UE1) may serve as an initiator, obtain federated learning task construction information, and determine at least one associated candidate participating device based on the federated learning task construction information.

[0181] Specifically, the federated learning task construction information may include business data features. In this example, the business data features may be image features or video features, and the federated learning task type may be a task to improve image or video clarity. User device 1 may determine at least one associated candidate participating device based on the image features or video features and the federated learning task type.

[0182] The determined candidate participating devices may include other user devices (such as UE2) that take photos or videos at locations corresponding to the image features or video features, public cloud devices that store photos or videos of scenic spots corresponding to the image features or video features, and other applications (App, Application), etc.

[0183] Step 2: User equipment 1 sends a request message for establishing federated learning to each candidate participating device. The request message may be a NAS message, an RRC message, a Sidelink message, or an application message.

[0184] For example, the message sent by the user equipment to the core network may be a NAS message, the message sent to the application module may be an application message, the message sent to other UEs may be a sidelink message, and the like.

[0185] The request message includes one or more of the federated learning task type and business data characteristics, and may also include information such as resources, computing power, energy consumption, etc. required to participate in the federated learning business.

[0186] Step 3: After receiving the request message for establishing federated learning, each candidate participating device may determine to accept or reject the request message.

[0187] Specifically, each candidate participating device may agree or reject the federated learning request message according to the content of the request message. For example, the candidate participating device may determine the relevance between the federated learning task to be constructed and itself according to one or more of the federated learning task type and business data characteristics. If the relevance is less than a preset value, the federated learning request message is rejected, otherwise the request message is accepted.

[0188] In some embodiments, the candidate participating device can determine whether it meets the resources, computing power, energy consumption, etc. required by the federated learning task in the message request based on its current resources, computing power, energy consumption and other parameters. If so, it accepts the request message, otherwise it rejects it.

[0189] In other embodiments, a corresponding single threshold value may also be set for each parameter in the resource, computing power, and energy consumption information required for the federated learning task in the message request. For example, when the resource utilization is greater than the preset resource utilization threshold, the request message is rejected, otherwise the request message is accepted. For another example, the candidate participating device can compare the total resource utilization of the currently used resources plus the resources required for the federated learning service with the resource utilization it can provide. If it is greater than the resource utilization it can provide, the request message is rejected, otherwise the request message is accepted.

[0190] In some embodiments, the user corresponding to the user device or application has the right to decide whether to participate in the federated learning task. If the user does not want to participate in the federated learning task, the user can also reject the request message through control operations.

[0191] Step 4: Each candidate participating device sends a confirmation message to the user device 1, which confirms that the candidate participating device accepts the request as the target participating device that finally participates in the federated learning. The user device 1 can construct a federated learning task based on the federated learning task construction information and perform the federated learning task with each target participating device.

[0192] like Figure 3 As shown, the confirmation message returned by the public cloud device and other applications is to accept the request, and the confirmation message returned by the user device 2 is to reject the request. At this time, the target participating devices determined are the public cloud device and other applications. User device 1 can construct a federated learning task based on the federated learning task construction information, and jointly participate in the execution of the federated learning task with the public cloud device and other applications.

[0193] Example 3: Type 3, when the federated learning system has a coordinator, and the federated learning type corresponding to the federated learning task to be constructed is vertical federated learning. The device interaction diagram of the federated system construction method provided in this application can be as follows: Figure 4 shown.

[0194] In step 1, base station 1 can act as an initiator and send a federated learning application message to a network management device as a coordinator. The federated learning application message may include federated learning task construction information. The network management device can determine at least one associated candidate participating device based on the federated learning task construction information.

[0195] Specifically, the federated learning task construction information may include the federated learning task type, data source information, and user information (such as user ID). The federated learning task type may be an analysis and prediction task for user device experience.

[0196] The network management device may determine at least one associated candidate participating device based on the type of federated learning task, and one or more of the data source information and the user information. The determined candidate participating devices may include a user equipment (UE) group corresponding to the data stored in base station 1, a base station having data of the group of UEs or serving the group of UEs (e.g., base station 1 and base station 2), and a core network element, etc.

[0197] In step 2, the network management device may send a request message for establishing federated learning (i.e., a federated learning request) to the candidate participating device. The request message may be a NAS message, an F1-AP message, or an RRC message. The request message may include one or more of the federated learning task type, data source information, and user information, and may also include information such as resources, computing power, and energy consumption required to participate in the federated learning service.

[0198] Step 3: After receiving the request message for establishing federated learning, each candidate participating device may determine to accept or reject the request message.

[0199] Specifically, each candidate participating device may agree or reject the federated learning request message according to the content of the request message. For example, the candidate participating device may determine the relevance between the federated learning task to be constructed and itself according to one or more of the federated learning task type and business data characteristics. If the relevance is less than a preset value, the federated learning request message is rejected, otherwise the request message is accepted.

[0200] In some embodiments, the candidate participating device can determine whether it meets the resources, computing power, energy consumption, etc. required by the federated learning task in the message request based on its current resources, computing power, energy consumption and other parameters. If so, it accepts the request message, otherwise it rejects it.

[0201] In other embodiments, a corresponding single threshold value may also be set for each parameter in the resource, computing power, and energy consumption information required for the federated learning task in the message request. For example, when the resource utilization is greater than the preset resource utilization threshold, the request message is rejected, otherwise the request message is accepted. For another example, the candidate participating device can compare the total resource utilization of the currently used resources plus the resources required for the federated learning service with the resource utilization it can provide. If it is greater than the resource utilization it can provide, the request message is rejected, otherwise the request message is accepted.

[0202] In some embodiments, the user corresponding to the user device or application has the right to decide whether to participate in the federated learning task. If the user does not want to participate in the federated learning task, the user can also reject the request message through control operations.

[0203] Step 4: Each candidate participating device sends a confirmation message to the network management device, which confirms that the candidate participating device accepts the request as the target participating device that ultimately participates in federated learning. The network management device can construct a federated learning task based on the federated learning task construction information and execute the federated learning task with each target participating device.

[0204] like Figure 2 As shown, the confirmation message returned by base station 1, the user device group and the core network element is to accept the request, and the confirmation message returned by base station 2 is to reject the request. At this time, the target participating devices determined are base station 1, the user device group and the core network element. The network management device can construct a federated learning task based on the federated learning task construction information, and the user device group and the core network element jointly participate in the device to execute the federated learning task.

[0205] Example 4: Type 4: When the federated learning system has a coordinator, and the federated learning type corresponding to the federated learning task to be constructed is horizontal federated learning, the device interaction diagram of the method for constructing a federated system provided in this application can be as follows: Figure 5 shown.

[0206] Step 1: User equipment 1 (UE1) can act as an initiator and send a federated learning application message to a core network element as a coordinator, wherein the federated learning application message includes federated learning task construction information. The core network element determines at least one associated candidate participating device based on the federated learning task construction information obtained in the federated learning application message.

[0207] Specifically, the federated learning task construction information may also include service data features. In this example, the service data features may be image features or video features, and the federated learning task type may be a task to improve image or video clarity. The core network element may determine at least one associated candidate participating device based on the image features or video features and the federated learning task type.

[0208] The determined candidate participating devices may include other user devices (such as UE1 and UE2) that take photos or videos at locations corresponding to the image features or video features, public cloud devices that store photos or videos of scenic spots corresponding to the image features or video features, and other applications (App, Application), etc.

[0209] Step 2: The core network element (i.e., the network element device on the network side) sends a request message for establishing federated learning to each candidate participating device. The request message may be a NAS message, an RRC message, a Sidelink message, or an application message.

[0210] For example, the message sent by the user equipment to the core network may be a NAS message, the message sent to the application module may be an application message, the message sent to other UEs may be a sidelink message, and the like.

[0211] The request message includes one or more of the federated learning task type and business data characteristics, and may also include information such as resources, computing power, energy consumption, etc. required to participate in the federated learning business.

[0212] Step 3: After receiving the request message for establishing federated learning, each candidate participating device may determine to accept or reject the request message.

[0213] Specifically, each candidate participating device may agree or reject the federated learning request message according to the content of the request message. For example, the candidate participating device may determine the relevance between the federated learning task to be constructed and itself according to one or more of the federated learning task type and business data characteristics. If the relevance is less than a preset value, the federated learning request message is rejected, otherwise the request message is accepted.

[0214] In some embodiments, the candidate participating device can determine whether it meets the resources, computing power, energy consumption, etc. required by the federated learning task in the message request based on its current resources, computing power, energy consumption and other parameters. If so, it accepts the request message, otherwise it rejects it.

[0215] In other embodiments, a corresponding single threshold value may also be set for each parameter in the resource, computing power, and energy consumption information required for the federated learning task in the message request. For example, when the resource utilization is greater than the preset resource utilization threshold, the request message is rejected, otherwise the request message is accepted. For another example, the candidate participating device can compare the total resource utilization of the currently used resources plus the resources required for the federated learning service with the resource utilization it can provide. If it is greater than the resource utilization it can provide, the request message is rejected, otherwise the request message is accepted.

[0216] In some embodiments, the user corresponding to the user device or application has the right to decide whether to participate in the federated learning task. If the user does not want to participate in the federated learning task, the user can also reject the request message through control operations.

[0217] Step 4: Each candidate participating device sends a confirmation message to the core network element, and the confirmation message is for the candidate participating device to accept the request as the target participating device that ultimately participates in federated learning.

[0218] like Figure 5As shown, the confirmation message returned by UE1, public cloud devices and other applications is to accept the request, and the confirmation message returned by user equipment 2 is to reject the request. At this time, the target participating devices determined are UE1, public cloud devices and other applications. The core network element can construct a federated learning task based on the federated learning task construction information, and the public cloud device and other applications jointly participate in the execution of the federated learning task.

[0219] The present application embodiment provides a federated learning system construction device, such as Figure 6 As shown, the federated learning system construction device 60 may include: a candidate device determination module 601, a learning request sending module 602, a target device determination module 603 and a federated learning module 604, wherein:

[0220] A candidate device determination module 601 is used to obtain federated learning task construction information and determine at least one associated candidate participating device based on the federated learning task construction information;

[0221] A learning request sending module 602 is used to send a federated learning request containing federated learning task construction information to each candidate participating device;

[0222] The target device determining module 603 is configured to receive reply information returned by at least one candidate participating device, and determine at least one target participating device from the at least one candidate participating device based on the reply information;

[0223] The federated learning module 604 is used to construct a federated learning task based on the federated learning task construction information, and execute the federated learning task with at least one target participating device.

[0224] The present application embodiment provides a federated learning system construction device, such as Figure 7 As shown, the federated learning system construction device 70 may include: a first device determination module 701, a first request sending module 702, a second device determination module 703 and a first federated learning module 704, wherein:

[0225] The first device determination module 701 is configured to receive federated learning application information, wherein the federated learning application information includes federated learning task construction information; determine at least one associated candidate participating device based on the federated learning task construction information, wherein the at least one candidate participating device includes a device that sends the federated learning application information;

[0226] A first request sending module 702 is used to send a federated learning request including federated learning task construction information to each candidate participating device;

[0227] The second device determination module 703 is configured to receive reply information returned by at least one candidate participating device, and determine at least one target participating device from the at least one candidate participating device based on the reply information;

[0228] The first federated learning module 704 is used to construct a federated learning task based on the federated learning task construction information, and execute the federated learning task with at least one target participating device.

[0229] The above-mentioned federated learning system construction device obtains the federated learning task construction information, and determines at least one associated candidate participating device based on the federated learning task construction information, and then sends a federated learning request containing the federated learning task construction information to each candidate participating device. After receiving the reply information returned by at least one candidate participating device, at least one target participating device can be determined from at least one candidate participating device based on the reply information, and then a federated learning task is constructed based on the federated learning task construction information, and the federated learning task is performed with at least one target participating device. The present application can flexibly and specifically determine participating devices according to business needs through the federated learning task construction information, thereby achieving a larger number of participating devices determined, and the data provided by each participating device is more in line with the training requirements of the federated learning task, and ultimately optimizes the training results of the federated learning task.

[0230] The device of the embodiments of the present application can execute the method provided by the embodiments of the present application, and the implementation principles are similar. The actions performed by each module in the device of each embodiment of the present application correspond to the steps in the method of each embodiment of the present application. For the detailed functional description of each module of the device, please refer to the description in the corresponding method shown in the previous text, which will not be repeated here.

[0231] An electronic device (computer device / equipment / system) is provided in an embodiment of the present application, including a memory, a processor and a computer program stored in the memory. The processor executes the above-mentioned computer program to implement the steps of the method for building a federated learning system. Compared with the related art, it can achieve: the ability to flexibly and specifically determine participating devices according to business needs, to achieve a larger number of participating devices determined, and the data provided by each participating device is more in line with the training requirements of the federated learning task, thereby ultimately optimizing the training results of the federated learning task.

[0232] In an alternative embodiment, an electronic device is provided, such as Figure 8 As shown, Figure 8The electronic device 4000 shown includes: a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, such as through a bus 4002. Optionally, the electronic device 4000 may also include a transceiver 4004, which may be used for data interaction between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present application.

[0233] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0234] The bus 4002 may include a path to transmit information between the above components. The bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0235] The memory 4003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store computer programs and can be read by a computer, without limitation herein.

[0236] The memory 4003 is used to store the computer program for executing the embodiment of the present application, and the execution is controlled by the processor 4001. The processor 4001 is used to execute the computer program stored in the memory 4003 to implement the steps shown in the above method embodiment.

[0237] The electronic devices include, but are not limited to, mobile terminals such as mobile phones, notebook computers, PADs, and fixed terminals such as digital TVs and desktop computers.

[0238] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps and corresponding contents of the aforementioned method embodiment can be implemented.

[0239] The embodiment of the present application also provides a computer program product, including a computer program, which can implement the steps and corresponding contents of the aforementioned method embodiment when executed by a processor.

[0240] The terms "first", "second", "third", "fourth", "1", "2", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than that shown or described in the drawings.

[0241] It should be understood that, although each operation step is indicated by arrows in the flowchart of the embodiment of the present application, the implementation order of these steps is not limited to the order indicated by the arrows. Unless clearly stated herein, in some implementation scenarios of the embodiment of the present application, the implementation steps in each flowchart can be performed in other orders according to demand. In addition, some or all of the steps in each flowchart may include multiple sub-steps or multiple stages based on actual implementation scenarios. Some or all of these sub-steps or stages may be executed at the same time, and each sub-step or stage in these sub-steps or stages may also be executed at different times respectively. In different scenarios of execution time, the execution order of these sub-steps or stages may be flexibly configured according to demand, and the embodiment of the present application does not limit this.

[0242] The above is only an optional implementation method for some implementation scenarios of the present application. It should be pointed out that for ordinary technicians in this technical field, without departing from the technical concept of the solution of the present application, other similar implementation methods based on the technical ideas of the present application are also within the protection scope of the embodiments of the present application.

Claims

1. A method for constructing a federated learning system, characterized in that: Applied to federated learning initiating devices, including: Acquire federated learning task construction information, and determine at least one associated candidate participating device based on the federated learning task construction information; Sending a federated learning request including the federated learning task construction information to each candidate participating device; receiving reply information returned by the at least one candidate participating device, and determining at least one target participating device from the at least one candidate participating device based on the reply information; Constructing a federated learning task based on the federated learning task construction information, and executing the federated learning task with the at least one target participating device; The determining, based on the federated learning task construction information, at least one associated candidate participating device comprises: Determine a corresponding federated learning type based on the federated learning task construction information; At least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information; the federated learning type is used to describe that business data features or user data features of the at least one candidate participating device are similar.

2. The method for constructing a federated learning system according to claim 1, characterized in that: The determining the corresponding federated learning type based on the federated learning task construction information includes: Extracting a federated learning task type from the federated learning task construction information; there is a corresponding relationship between the federated learning task type and the federated learning type; The federated learning type corresponding to the federated learning task construction information is determined based on the corresponding relationship and the federated learning task type.

3. The method for constructing a federated learning system according to claim 2, wherein: The types of federated learning include horizontal federated learning; When the federated learning type is the horizontal federated learning, determining at least one associated candidate participating device based on the federated learning type and the federated learning task construction information includes: The business data features in the federated learning task construction information are extracted, and the devices having the business data features are used as the candidate participating devices.

4. The method for constructing a federated learning system according to claim 2, characterized in that: The types of federated learning include vertical federated learning; When the federated learning type is the vertical federated learning, determining at least one associated candidate participating device based on the federated learning type and the federated learning task construction information includes: The user data features in the federated learning task construction information are extracted, and the devices corresponding to the user data features are used as the candidate participating devices.

5. The method for constructing a federated learning system according to claim 1, characterized in that: The method further comprises: Sending federated learning application information to a federated learning coordination device, wherein the federated learning application information includes the federated learning task construction information; Receive a federated learning request sent by a federated learning coordination device, and return reply information to the federated learning coordination device for the federated learning request; if the reply information is confirmed, then perform the federated learning task as one of the target participating devices together with the federated learning coordination device and other target participating devices.

6. The method for constructing a federated learning system according to any one of claims 1 to 5, characterized in that: The federated learning request also includes at least one of the following: resource information, computing power information, and energy consumption information required to participate in the federated learning.

7. The method for constructing a federated learning system according to claim 6, characterized in that: When each candidate participating device receives the federated learning request, it performs the following operations: Determine a degree of relevance to the federated learning request according to the required resource information, at least one of the computing power information and the energy consumption information in the federated learning request, and the federated learning task construction information; If the correlation degree is greater than or equal to a preset threshold, the reply information returned for the federated learning request is confirmation; If the correlation level is less than a preset threshold, the response information returned for the federated learning request is rejection.

8. A method for constructing a federated learning system, characterized in that: Applied to federated learning coordination devices, including: Receiving federated learning application information, wherein the federated learning application information includes federated learning task construction information; Determine at least one associated candidate participating device based on the federated learning task construction information, wherein the at least one candidate participating device includes a device that sends the federated learning application information; Sending a federated learning request including the federated learning task construction information to each candidate participating device; Receive reply information returned by the at least one candidate participating device, determine at least one target participating device from the at least one candidate participating device based on the reply information; construct a federated learning task based on the federated learning task construction information, and execute the federated learning task together with the at least one target participating device; The determining, based on the federated learning task construction information, at least one associated candidate participating device comprises: Determine a corresponding federated learning type based on the federated learning task construction information; At least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information; the federated learning type is used to describe that business data features or user data features of the at least one candidate participating device are similar.

9. The method for constructing a federated learning system according to claim 8, characterized in that: The determining the corresponding federated learning type based on the federated learning task construction information includes: Extracting a federated learning task type from the federated learning task construction information; there is a corresponding relationship between the federated learning task type and the federated learning type; The federated learning type corresponding to the federated learning task construction information is determined based on the corresponding relationship and the federated learning task type.

10. The method for constructing a federated learning system according to claim 9, characterized in that: The types of federated learning include horizontal federated learning; When the federated learning type is the horizontal federated learning, determining at least one associated candidate participating device based on the federated learning type and the federated learning task construction information includes: The business data features in the federated learning task construction information are extracted, and the devices having the business data features are used as the candidate participating devices.

11. The method for constructing a federated learning system according to claim 9, characterized in that: The types of federated learning include vertical federated learning; When the federated learning type is the vertical federated learning, determining at least one associated candidate participating device based on the federated learning type and the federated learning task construction information includes: The user data features in the federated learning task construction information are extracted, and the devices corresponding to the user data features are used as the candidate participating devices.

12. A federated learning system construction device, characterized in that: include: A candidate device determination module, configured to obtain federated learning task construction information and determine at least one associated candidate participating device based on the federated learning task construction information; A learning request sending module, used to send a federated learning request containing the federated learning task construction information to each candidate participating device; a target device determining module, configured to receive reply information returned by the at least one candidate participating device, and determine at least one target participating device from the at least one candidate participating device based on the reply information; a federated learning module, configured to construct a federated learning task based on the federated learning task construction information, and execute the federated learning task with the at least one target participating device; The determining, based on the federated learning task construction information, at least one associated candidate participating device comprises: Determine a corresponding federated learning type based on the federated learning task construction information; At least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information; the federated learning type is used to describe that business data features or user data features of the at least one candidate participating device are similar.

13. A federated learning system construction device, characterized in that: include: A first device determination module is configured to receive federated learning application information, wherein the federated learning application information includes federated learning task construction information; determine at least one associated candidate participating device based on the federated learning task construction information, wherein the at least one candidate participating device includes a device that sends the federated learning application information; A first request sending module, configured to send a federated learning request including the federated learning task construction information to each candidate participating device; A second device determination module, configured to receive reply information returned by the at least one candidate participating device, and determine at least one target participating device from the at least one candidate participating device based on the reply information; A first federated learning module, configured to construct a federated learning task based on the federated learning task construction information, and execute the federated learning task with the at least one target participating device; The determining, based on the federated learning task construction information, at least one associated candidate participating device comprises: Determine a corresponding federated learning type based on the federated learning task construction information; At least one associated candidate participating device is determined based on the federated learning type and the federated learning task construction information; the federated learning type is used to describe that business data features or user data features of the at least one candidate participating device are similar.

14. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 11.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for constructing a federated learning system according to any one of claims 1 to 11 are implemented.

Images