Distributed training method, system and device, storage medium and program product

By introducing distributed training methods in federated learning, using the scheduling platform to generate and send target scheduling information, the training equipment can flexibly schedule multiple algorithm components, solving the problems of low training efficiency and difficult code maintenance in the existing technology, and realizing the Efficient joint training of algorithm components.

CN118963977AActive Publication Date: 2024-11-15LINGYANG INTELLIGENT TECHNOLOGY CO LTD +1
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202411464800.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-21
Publication Date
2024-11-15
Estimated Expiration
2044-10-21

AI Technical Summary

Technical Problem

The existing federated learning technology cannot flexibly schedule multiple training algorithms and security algorithms, resulting in inefficient training and difficulty in maintaining code.

Method used

Provide a distributed training method, which receives user configuration information through the scheduling platform, generates target scheduling information of the training device, and sends it to the training device, so that it can schedule and combine multiple algorithm components to realize flexible scheduling of algorithm components. .

Benefits of technology

It improves training efficiency, reduces the need for code writing and maintenance, and realizes flexible scheduling and efficient joint training of multiple algorithms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118963977A_ABST
    Figure CN118963977A_ABST
Patent Text Reader

Abstract

The invention discloses a distributed training method, system and device, a storage medium and a program product, and the method comprises the steps that a scheduling platform receives configuration information submitted by a user, and generates target scheduling information of each training device according to initial scheduling information of at least two training devices participating in joint training included in the configuration information; and sending the target scheduling information of each training device to the corresponding training device, so that each training device schedules a plurality of pre-maintained algorithm components according to the corresponding target scheduling information, and performing joint training by using the scheduled target algorithm components. According to the invention, flexible scheduling of multiple algorithms in distributed training is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and specifically to a distributed training method, system, device, storage medium and program product. Background Art

[0002] Distributed training is the joint training of machine learning models on multiple training nodes. Federated learning is one of the typical distributed training methods, which involves multiple algorithms, such as training algorithms and security algorithms. At present, in the process of federated training, training is often carried out based on fixed training algorithms and fixed security algorithms according to training requirements, and it is impossible to flexibly schedule multiple algorithms such as multiple training algorithms and multiple security algorithms. Summary of the invention

[0003] In order to solve the problem that multiple algorithms cannot be flexibly scheduled in the above-mentioned related technologies, the present application provides a distributed training method, system, device, storage medium and program product.

[0004] A first aspect of the present application provides a distributed training method, which is applied to a scheduling platform. The method includes: receiving configuration information submitted by a user; the configuration information includes initial scheduling information of at least two training devices participating in the joint training, and the initial scheduling information includes partial scheduling information of multiple algorithm components; Generate target scheduling information of the training device according to the initial scheduling information of the training device; the target scheduling information includes complete scheduling information of multiple algorithm components; The target scheduling information of the training device is sent to the training device respectively. The target scheduling information is used to schedule multiple algorithm components based on the target scheduling information, and joint training is performed based on the scheduled target algorithm components. The training device maintains multiple algorithm components in advance.

[0005] A second aspect of the present application provides a distributed training method, applied to a training device, the method comprising: receiving target scheduling information sent by the scheduling platform; the target scheduling information is generated by the scheduling platform after receiving the configuration information submitted by the user, based on the initial scheduling information of at least two training devices participating in the joint training included in the configuration information; the initial scheduling information includes partial scheduling information of multiple algorithm components, and the target scheduling information includes complete scheduling information of multiple algorithm components; According to the target scheduling information, scheduling processing is performed on multiple algorithm components maintained in advance to obtain a target algorithm component; Joint training with target algorithm components.

[0006] A third aspect of the present application provides a distributed training device, which is applied to a scheduling platform. The device includes: A receiving module, configured to receive configuration information submitted by a user, wherein the configuration information includes initial scheduling information of at least two training devices participating in the joint training; the initial scheduling information includes partial scheduling information of multiple algorithm components; A generation module, used for generating target scheduling information of the training device according to the initial scheduling information of the training device; the target scheduling information includes complete scheduling information of multiple algorithm components; The sending module is used to send the target scheduling information of the training device to the training device respectively. The target scheduling information is used to schedule multiple algorithm components based on the target scheduling information, and perform joint training based on the scheduled target algorithm components. The training device maintains multiple algorithm components in advance.

[0007] A fourth aspect of the present application provides a distributed training device, applied to a training device, the device comprising: A receiving module, used to receive target scheduling information sent by the scheduling platform; the target scheduling information is generated by the scheduling platform after receiving the configuration information submitted by the user, based on the initial scheduling information of at least two training devices participating in the joint training included in the configuration information; the initial scheduling information includes partial scheduling information of multiple algorithm components, and the target scheduling information includes complete scheduling information of multiple algorithm components; The scheduling module is used to schedule the various algorithm components maintained in advance according to the target scheduling information to obtain the target algorithm component; A training module is used for joint training with target algorithm components.

[0008] A fifth aspect of the present application provides a distributed training system, the system comprising: a scheduling platform and a plurality of training devices; The configuration and scheduling platform is used to receive configuration information submitted by a user, wherein the configuration information includes initial scheduling information of at least two training devices participating in the joint training among the multiple training devices; generate target scheduling information of the training device according to the initial scheduling information of the training device; and send the target scheduling information of the training device to the training device respectively; the initial scheduling information includes partial scheduling information of multiple algorithm components, and the target scheduling information includes complete scheduling information of multiple algorithm components; The training device is used to receive the target scheduling information sent by the scheduling platform; according to the target scheduling information, perform scheduling processing on multiple algorithm components maintained in advance to obtain the target algorithm component; and perform joint training using the target algorithm component.

[0009] The sixth aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in the first aspect above, or to implement the method described in the second aspect above.

[0010] The seventh aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement the method as described in the first aspect above, or to implement the method as described in the second aspect above.

[0011] An eighth aspect of the present application provides a computer program product, including a computer program, wherein the computer program is executed by a processor to implement the method described in the first aspect above, or to implement the method described in the second aspect above.

[0012] Based on the distributed training method provided in the first aspect above, the present application has at least the following beneficial effects or advantages: In an embodiment of the present application, when the scheduling platform receives the configuration information input by the user, the target scheduling information of each training device is generated according to the initial scheduling information of at least two training devices participating in the joint training included in the configuration information; and the target scheduling information of each training device is sent to the corresponding training device respectively, so that each training device schedules the pre-maintained multiple algorithm components according to the corresponding target scheduling information, and uses the scheduled target algorithm components for joint training. Thus, by pre-constructing multiple algorithm components and maintaining the multiple algorithm components by each training device, and providing a configuration function through the scheduling platform, the scheduling platform can determine the target scheduling information according to the initial scheduling information input by the user, thereby sending the target scheduling information to the training device to schedule the multiple algorithm components maintained by the training device. It can be seen that the flexible scheduling of multiple algorithms is realized in this training method. In addition, since the training device can directly perform joint training based on the scheduled target algorithm component without the need to perform code writing and other operations based on the corresponding training requirements, the training efficiency is greatly improved.

[0013] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A schematic diagram of an application scenario of a distributed training method provided in an embodiment of the present application; Figure 2 A first flow chart of a distributed training method provided in an embodiment of the present application; Figure 3A second flow chart of a distributed training method provided in an embodiment of the present application; Figure 4 A third flow chart of a distributed training method provided in an embodiment of the present application; Figure 5 A fourth flow chart of a distributed training method provided in an embodiment of the present application; Figure 6 A fifth flow chart of a distributed training method provided in an embodiment of the present application; Figure 7 A schematic diagram of the training process of the interactive training mode provided in an embodiment of the present application; Figure 8 A schematic diagram of the training process of the centralized training mode provided in an embodiment of the present application; Fig. 9 A first structural schematic diagram of a distributed training device provided in an embodiment of the present application; Fig.10 A second structural schematic diagram of a distributed training device provided in an embodiment of the present application; Fig.11 A schematic diagram of the composition of a distributed training system provided in an embodiment of the present application; Fig.12 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application; Fig.13 A schematic diagram of the structure of a storage medium provided in an embodiment of the present application. DETAILED DESCRIPTION

[0015] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0016] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following explanations: Federated Learning (FL): A machine learning technology that allows multiple training devices to jointly train models without sharing data. It can improve model accuracy and training efficiency while protecting data privacy.

[0017] Differential Privacy (DP): A data privacy protection technology that protects the privacy of data by adding noise to the data.

[0018] Homomorphic Encryption (HE): An encryption technology that can encrypt and decrypt while maintaining the structure and original meaning of data. It can improve data security while protecting data privacy.

[0019] Multi-party secure computing (MPC) is a data privacy protection technology that can split a piece of data into multiple copies. Each copy of the data cannot be restored to the original data, but after separate calculations, the result of centralized calculation can be achieved. The decrypted result can be used for federated learning.

[0020] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0021] It should be understood that, although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determination", etc.

[0022] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0023] Federated learning, as a distributed machine learning method that protects data privacy, can perform collaborative training of models without leaving the local device. In other words, multiple algorithms such as security algorithms and training algorithms are used in the federated learning process. Although federated learning provides stronger protection for data privacy and security, the existing federated learning does not support the flexible scheduling of multiple algorithms such as multiple security algorithms and multiple training algorithms. That is, when R&D personnel obtain training requirements, they often write corresponding codes according to the security algorithms and training algorithms specified in the training requirements. The specified security algorithms and training algorithms are coupled together and cannot be scheduled. However, once the training requirements change, code adjustments are required, and code adjustments often take a long time, thereby reducing training efficiency. Furthermore, different training requirements may have the same security algorithm or the same training algorithm, and the corresponding code is written for each training requirement, which not only has a low code reuse rate, but also increases the amount of code required for maintenance.

[0024] Based on this, the embodiment of the present application provides a distributed training method. Figure 1 Schematic diagram of an application scenario of a distributed training method provided in an embodiment of the present application. Figure 1 As shown, the scenario includes a scheduling platform and multiple training devices. The scheduling platform can be set in the user's terminal device or in the server, and the scheduling platform is connected to each training device in communication. The training device can be a terminal device or a server, and each training device maintains a variety of algorithm components, which can be set in the training device or not, for example, set in the cloud. Among them, the terminal device can be a mobile phone, a tablet computer, a desktop computer, a portable notebook, etc. The server can be a physical server, a cloud server, etc. When the scheduling platform is set in the terminal device, the scheduling platform can be an independent application (Application, referred to as App), or a small program set in the application, or a web application, etc. When the scheduling platform is set on the server side, the user's terminal device can be provided with a related client that communicates with the scheduling platform, and the client can be an independent application, or a small program set in the application, or a web application, etc. The client is used to send the configuration information entered by the user to the scheduling platform.

[0025] Figure 1 In the example, the scheduling platform is set in the user's desktop computer and there are two training devices participating in the joint training. It should be understood that Figure 1The illustration is only a schematic representation of the application scenario of the distributed training method involved in the present application, and does not constitute a limitation on the technical solution of the present application. In some other embodiments, the application scenario of the distributed training method involved in the present application may also include more or fewer components.

[0026] Specifically, in Figure 1 In the embodiment, the scheduling platform may include a human-computer interaction interface, which includes a login interface, a configuration interface, etc. Among them, the login interface is used for users to log in to the scheduling platform, and the configuration interface is used for users to edit the configuration information of distributed training. The scheduling platform can maintain the login information of at least one user, the algorithms to which the user has access rights, the device information of each training device to which the user has access rights, the sample information of the training samples corresponding to each training device, the generation strategy of the target scheduling information, the sending strategy of the target scheduling information, etc. Among them, the device information includes the device identification, the communication address of the device, etc. The sample information includes the label information, the library information of the database where the training sample is located, etc. The generation strategy of the target scheduling information is used to instruct the scheduling platform to generate the target scheduling information of each training device according to the initial scheduling information of at least two training devices participating in the joint training included in the configuration information input by the user. The sending strategy of the target scheduling information is used to instruct the scheduling platform to send the generated target scheduling information to the corresponding training device participating in the joint training. Each training device maintains a plurality of pre-built algorithm components, scheduling strategies of algorithm components, training strategies, etc. The multiple algorithm components include multiple security components in a security component set, multiple training components in a training component set, and multiple verification components in a verification component set. Each security component is used to execute a security algorithm, each training component is used to execute a training algorithm, and each verification component is used to execute a verification algorithm. The scheduling strategy is used to instruct the training device to schedule the multiple algorithm components that are maintained in advance according to the received target scheduling information. The training strategy is used to instruct the training device to perform joint training based on the scheduled target algorithm components.

[0027] It can be seen that by pre-building multiple algorithm components and having each training device maintain the multiple algorithm components, and providing a configuration function through the scheduling platform, the scheduling platform can determine the target scheduling information based on the initial scheduling information input by the user, and then send the target scheduling information to the training device to schedule the multiple algorithm components maintained by the training device. It can be seen that this training method realizes flexible scheduling of multiple algorithms. In addition, since the training device can directly perform joint training based on the scheduled target algorithm component without the need to perform code writing and other operations based on the corresponding training requirements, the training efficiency is greatly improved.

[0028] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The several specific embodiments listed can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0029] Figure 2 A flowchart of a distributed training method provided in an embodiment of the present application, Figure 2 The method shown can be Figure 1 The scheduling platform in is executed. Figure 2 As shown, the method includes the following steps 101-103.

[0030] Step 101, receiving configuration information submitted by a user; the configuration information includes initial scheduling information of at least two training devices participating in the joint training, and the initial scheduling information includes partial scheduling information of multiple algorithm components.

[0031] Step 102: Generate target scheduling information for each training device based on the initial scheduling information for each training device. The target scheduling information includes complete scheduling information for multiple algorithm components.

[0032] Step 103, the target scheduling information of each training device is sent to the corresponding training device respectively, the target scheduling information is used to schedule multiple algorithm components based on the target scheduling information, and the scheduled target algorithm components are used for joint training, and the training device maintains multiple algorithm components in advance.

[0033] In some embodiments, the scheduling platform can be set in the user's terminal device. Accordingly, when the user has a distributed training requirement, the user can operate the scheduling platform in his terminal device to input configuration information and submit it. In response to the user's submission operation, the scheduling platform obtains the configuration information submitted by the user; and generates target scheduling information for each training device according to the initial scheduling information of at least two training devices participating in the joint training included in the configuration information, and sends the target scheduling information to the corresponding training device.

[0034] In other embodiments, the scheduling platform may be provided at the server end, and a client that is in communication with the scheduling platform may be provided in the user's terminal device. Accordingly, when the user has a distributed training requirement, the user may operate the client in his terminal device to input the configuration information and submit it. In response to the user's submission operation, the client obtains the configuration information submitted by the user and sends it to the scheduling platform. The scheduling platform generates the target scheduling information of each training device according to the initial scheduling information of each of the at least two training devices participating in the joint training included in the received configuration information, and sends the target scheduling information to the corresponding training device.

[0035] When the training device receives the target scheduling information sent by the scheduling platform, it schedules and processes the pre-maintained multiple algorithm components according to the target scheduling information to obtain the target algorithm component; and uses the target algorithm component for joint training. Among them, any algorithm component among the multiple algorithm components is pre-coded according to the algorithm executed by the any algorithm component, and the written code is encapsulated. It can be understood that since the same algorithm component can be used in different training scenarios without writing corresponding codes based on each training scenario, the code reuse rate is improved and the amount of code required to be maintained is reduced.

[0036] The above-mentioned initial scheduling information may include the device identification, security algorithm, training algorithm, model verification algorithm, sample information, etc. of at least two training devices participating in the joint training. The target scheduling information may add information such as the training mode and the communication address of the device on the basis of the initial scheduling information. That is, the initial scheduling information includes partial scheduling information of multiple algorithm components, and the target scheduling information includes complete scheduling information of multiple algorithm components. Among them, the security algorithm may include a first security algorithm and a second security algorithm, the first security algorithm refers to the algorithm used when data is exchanged between the training devices participating in the joint training, and the second security algorithm refers to the algorithm used for the intersection of the privacy set of the training samples. The training algorithm is the algorithm used in the model training. The model verification algorithm refers to the algorithm used to verify the performance of the model. The first security algorithm may be any one of the secp256k1 (Secure Elliptic Curve 256k1) elliptic curve, Curve25519 elliptic curve, Ed25519 (Edwards-curve Digital Signature Algorithm (EdDSA) over Curve25519) elliptic curve, etc. The second security algorithm may be any one of homomorphic encryption, differential privacy, multi-party secure computing, etc. The training algorithm may include any one of logistic regression, linear regression, decision trees, random forest, etc. The model validation algorithm includes any one of cross-validation, recall, mean squared error (MSE), etc.

[0037] In the distributed training process, at least two devices participating in the joint training can be divided into a master device (also called a host device or host party) and a slave device (also called a slave device or slave party). The master device holds a first training sample with a label, and the slave device holds a second training sample without a label. In the joint training process, the operations performed by the master device and the slave device are different, so the scheduling platform generates the target configuration information of each training device participating in the joint training based on the initial scheduling information of each training device participating in the joint training included in the configuration information input by the user. The operations performed by the master device and the slave device during the training process can be found in the relevant description below.

[0038] In the distributed training method provided by the embodiment of the present application, when the scheduling platform receives the configuration information submitted by the user, the target scheduling information of each training device is generated according to the initial scheduling information of at least two training devices participating in the joint training included in the configuration information; and the target scheduling information of each training device is sent to the corresponding training device respectively, so that each training device schedules the pre-maintained multiple algorithm components according to the corresponding target scheduling information, and uses the scheduled target algorithm components for joint training. Thus, by pre-constructing multiple algorithm components and maintaining the multiple algorithm components by each training device, and providing a configuration function through the scheduling platform, the scheduling platform can determine the target scheduling information according to the initial scheduling information input by the user, so as to send the target scheduling information to the training device to schedule the multiple algorithm components maintained by the training device. It can be seen that the flexible scheduling of multiple algorithms is realized in this training method. In addition, since the training device can directly perform joint training based on the scheduled target algorithm component without the need to perform code writing and other operations based on the corresponding training requirements, the training efficiency is greatly improved.

[0039] In order to facilitate the management of the training devices and algorithms with access rights of each user, in some implementations, each user may first register on the scheduling platform, and after successful registration, log in to the successfully registered account to add or delete their own devices and algorithms. Specifically, the method may further include the following steps A1 and A2: Step A1, obtain the user's configuration item processing request; the configuration item processing request is used to request new addition processing and / or deletion processing of the target configuration item; the target configuration item includes at least one of a training device, a training sample, a security algorithm, a training algorithm, and a model verification algorithm.

[0040] Step A2: According to the configuration item processing request, the target configuration item is processed accordingly.

[0041] In some embodiments, the configuration item processing request is used to request the training device to be processed. Accordingly, after the user successfully logs in to the scheduling platform, the user can edit the device information of at least one training device to be added, the sample information of the training sample corresponding to the training device, etc. in the configuration item adding interface of the scheduling platform and submit it to send the configuration item processing request to the scheduling platform. Accordingly, the scheduling platform responds to the user's submission operation, obtains the configuration item processing request for submission based on the adding interface, and establishes a first association relationship between the device information and sample information included in the configuration item processing request and the user information of the user (such as the account information of the successful registration, the user identification, etc.), and saves the first association relationship to indicate that the user has the access right to the device corresponding to the device information. Among them, the device information may include a device identification, a communication address, etc. The communication address is used to locate the training device and communicate data with the training device. The communication address may be an IP (Internet Protocol) address, a MAC (Media Access Control) address, etc. The sample information may include the library information of the database where the training sample is located, such as the access address of the database, the library identification of the database, etc. When the training sample corresponding to the training device carries a label, the sample information may also include label information. It should be noted that each training device can correspond to at least one sample information, that is, each training device can access at least one training sample in the database. The training devices to which the user has access rights include training devices added by the user's operation scheduling platform, training devices added by other users' operation scheduling platforms and granted to the user access rights, etc.

[0042] Considering that in actual applications, the user may no longer use at least one training device to which he has access rights for distributed training due to various reasons, at this time, the user can also operate the scheduling platform to delete the corresponding training device in the scheduling platform. That is, in some embodiments, the configuration item processing request is used to request the deletion of the training device. Accordingly, after the user successfully logs in to the scheduling platform, he can edit (for example, check the multiple training devices displayed in the deletion interface) the device information of at least one training device to be deleted in the configuration item deletion interface of the scheduling platform and submit it to send a configuration item processing request to the scheduling platform. Accordingly, the scheduling platform responds to the user's submission operation, obtains the configuration item processing request submitted by the user based on the deletion page, obtains the corresponding target association relationship in the saved first association relationship according to the device information included in the configuration item processing request, and deletes the target association relationship.

[0043] When the target configuration items are training samples, security algorithms, training algorithms, and model verification algorithms, the processes of adding and deleting them are similar to those of the above-mentioned training devices. Please refer to the above-mentioned relevant descriptions and will not be listed one by one here.

[0044] It should be noted that when the target configuration items include multiple ones (such as training equipment and security algorithms, etc.), some of the configuration items (such as training equipment, etc.) can be added, and another part of the configuration items (such as security algorithms, etc.) can be deleted; when there is only one target configuration item (such as training equipment), the target configuration item is added or deleted. When the target configuration item is a training sample, if a new training sample is added, the configuration item processing request may include the library information of the database where the training sample is located and the device information of the training device corresponding to the training sample; if the training sample is deleted, the configuration item processing request may include the library information of the database where the training sample is located. When the target configuration item is a security algorithm, a training algorithm, or a model verification algorithm, the configuration item processing request may include at least one of the algorithm name, algorithm identifier, etc. of the corresponding algorithm.

[0045] It should be further pointed out that when the target configuration item is a training device or a training sample, the configuration item processing request can also be used to request modification of the target configuration item. For example, the communication address of the training device may be modified because the geographical location of the training device has changed, or the library information of the database may be modified because the training sample has been migrated to another database.

[0046] Therefore, based on the scheduling platform, users are provided with the addition and deletion processing of training equipment, training samples and various algorithms, which not only provides a guarantee for users to flexibly schedule multiple algorithms and multiple training equipment, but also enables users to conveniently manage their training equipment and multiple algorithms. At the same time, it also facilitates the scheduling platform to manage the training equipment, algorithms, etc. to which each user has access rights.

[0047] In order to flexibly and accurately schedule multiple algorithms and multiple devices to which the user has access rights during distributed training, in some implementations, after the user successfully logs into the scheduling platform, the user submits configuration information including initial scheduling information based on the configuration interface of the scheduling platform. Specifically, Figure 3 As shown, the above-mentioned step 101 may also include the following steps 1001 and 1002.

[0048] Step 1001: If it is determined that the login information submitted by the user to the scheduling platform is valid, then the device information of multiple training devices and the algorithm information of multiple algorithms to which the user has access rights are obtained.

[0049] Specifically, before performing the configuration operation, the user may first open the scheduling platform. In response to the user's opening operation, the scheduling platform displays a login interface and obtains the login information submitted by the user based on the login interface. The obtained login information is matched with the pre-saved login information. If the match is successful, it is determined that the login information submitted by the user to the scheduling platform is valid. And according to the user information of the user, the associated multiple device information and multiple algorithm information are obtained in the saved first association relationship.

[0050] Step 1002: display a configuration interface according to the acquired multiple device information and multiple algorithm information, where the configuration interface is used to display the multiple device information and multiple algorithm information.

[0051] Among them, multiple device information and multiple algorithm information can be set in the configuration interface in a flat manner, or in the form of a drop-down menu. The specific form of the configuration interface is not specifically limited in this application.

[0052] Corresponding to step 1001 and step 1002, as Figure 3 As shown, step 101 may include the following steps 1011: Step 1011, receiving configuration information submitted by a user based on a configuration interface; the configuration information includes initial scheduling information of at least two training devices participating in the joint training.

[0053] By providing a configuration interface to the user, the convenience of the user's configuration operation is improved, and it is also conducive to accurately obtaining the user's configuration information.

[0054] After obtaining the user configuration information, the target scheduling information of each training device participating in the joint training can be generated according to the configuration information. Figure 4 As shown, step 102 may include the following steps 1021 and 1022 .

[0055] Step 1021, determining the public configuration information and private configuration information corresponding to each training device according to the initial scheduling information of each training device.

[0056] Considering that in the distributed training process, the security algorithm, model verification algorithm and training mode corresponding to the master device and the slave device are usually the same. Based on this, in order to improve the efficiency and accuracy of determining the public configuration information, in some embodiments, in step 1021, the public configuration information corresponding to each training device is determined according to the initial scheduling information of each training device, which may include: obtaining the security algorithm, training algorithm and model verification algorithm from the initial scheduling information; determining the training mode according to the security algorithm and the training algorithm; determining the security algorithm, model verification algorithm and training mode as the public configuration information corresponding to each training device. In other words, the public configuration information is the same configuration information of at least two training devices participating in the joint training.

[0057] Wherein, determining the training mode according to the security algorithm and the training algorithm may include: according to the security algorithm and the training algorithm, in the second association relationship among the security algorithm, the training algorithm and the training mode, obtaining the associated training mode. As an example, if the security algorithm is a homomorphic encryption algorithm and the training algorithm is a logistic regression algorithm, then determining the training mode is an interactive training mode.

[0058] Considering that in the distributed training process, the training samples, communication addresses, etc. corresponding to the master device and the slave device are often different. Based on this, in order to ensure the accuracy of the determined private information, in some embodiments, in step 1021, the private configuration information corresponding to each training device is determined according to the initial scheduling information of each training device, which may include: determining the training algorithm and the communication address corresponding to each training device according to the initial scheduling information of each training device; determining the sample information of the training samples corresponding to each training device; for any training device in each training device, if the training device is the master device of the joint training, the training algorithm, the sample information corresponding to the training device, and the communication address of the slave device of the training device are determined as the private configuration information corresponding to the training device; if the training device is the slave device of the joint training, the sample information of the training device and the communication address of the master device of the training device are determined as the private configuration information corresponding to the training device. In other words, the private configuration information is different configuration information of at least two training devices participating in the joint training.

[0059] Specifically, the training algorithm, the device identification of each training device participating in the joint training, and the sample information corresponding to each training device participating in the joint training are obtained from the initial scheduling information; for each training device participating in the joint training, the associated communication address is obtained in the first stored association relationship according to the device identification of the training device, and the acquired communication address is determined as the communication address corresponding to the training device; and, if it is determined that the sample information corresponding to the training device includes label information, it is determined that the training device is the master device of the joint training, and if it is determined that the sample information corresponding to the training device does not include label information, it is determined that the training device is the slave device of the joint training. In the case where the training device is the master device, the training algorithm, the sample information corresponding to the training device, and the communication address of the slave device are determined as the private configuration information corresponding to the training device. In the case where the training device is a slave device, the sample information of the training device and the communication address of the master device are determined as the private configuration information corresponding to the training device.

[0060] For example, the training devices participating in the joint training include training device 1 and training device 2, where training device 1 is the master device and training device 2 is the slave device. The training algorithm, the sample information corresponding to training device 1, and the communication address of training device 2 are determined as the private configuration information corresponding to training device 1; the sample information corresponding to training device 2 and the communication address of training device 1 are determined as the private configuration information corresponding to training device 2.

[0061] Step 1022: Generate target scheduling information for each training device based on the public configuration information and private configuration information corresponding to each training device.

[0062] Specifically, for each training device, the public configuration information and private configuration information of the training device are determined as the target scheduling information of the training device.

[0063] That is, sending the target scheduling information of each training device to the corresponding training device in step 103 may include: for any one of the training devices, sending the public configuration information of the training device and the private configuration information corresponding to the training device to the training device.

[0064] By determining the public configuration information and private configuration information corresponding to each training device participating in the joint training, and generating the target configuration information of the training device based on the public configuration information and private configuration information corresponding to each training device, the accuracy of the target configuration information is guaranteed, thereby providing a guarantee for the subsequent accurate scheduling algorithm components.

[0065] Considering that in actual applications, different training samples in the same database may not be identical in structure, form, etc., in order to meet the training requirements. In some implementations, the user may also send a processing request for the training sample to the corresponding training device through the scheduling platform. Specifically, the method may also include the following steps B1 and B2: Step B1: If processing information of the target training sample submitted by the user is obtained, the target training device corresponding to the target training sample is determined according to the processing information.

[0066] The processing information may include access information of the target training sample, and the access information may be library information of the database where the target training sample is located, and the library information may be a database identifier or an access address of the database. Accordingly, when the scheduling platform obtains the processing information submitted by the user, it obtains the associated device information from the saved first association relationship according to the library information included in the processing information, and determines the training device corresponding to the obtained device information as the target training device corresponding to the target training sample.

[0067] Step B2: Send a processing request to the target training device according to the processing information; the processing request is used to request the target training device to process the target training sample according to the target structure standard to obtain a standardized training sample.

[0068] Among them, the target structure standard can be a user-specified structure standard included in the processing information, or it can be a preset structure standard. The target structure standard can include setting the missing data of the first feature of the target training sample as the target data, expanding the second feature of the target training sample from the first form to the second form, etc. For example, the target structure standard is to set the missing data of the feature "address" in the training sample to address 1; for another example, the target structure standard is to expand the feature "time" in the training sample from "yyyy.mm.dd" to "yyyy year mm month dd day". For the target structure standard, there is no specific limitation in this application, and it can be set as needed in actual applications.

[0069] It should be noted that the training samples can be various types of samples such as images, texts, audio and video, etc., and this application does not make any specific limitation on this.

[0070] By providing the processing function of training samples through the scheduling platform, users only need to operate the scheduling platform to send processing requests to the corresponding target training devices, without the need for users to manually process the target training samples, which greatly improves the processing efficiency of training sample standardization.

[0071] Corresponding to the aforementioned embodiment of the distributed training method, the present application also provides an embodiment of a distributed training method. Figure 5is a flow chart of a distributed training method according to an exemplary embodiment. Figure 5 The method shown can be Figure 1 The training equipment in Figure 5 As shown, the method includes the following steps 201-203.

[0072] Step 201, receiving the target scheduling information sent by the scheduling platform; the target scheduling information is generated by the scheduling platform after receiving the configuration information submitted by the user, based on the initial scheduling information of at least two training devices participating in the joint training included in the configuration information; the initial scheduling information includes partial scheduling information of multiple algorithm components, and the target scheduling information includes complete scheduling information of multiple algorithm components.

[0073] Step 202, according to the target scheduling information, scheduling processing is performed on the multiple algorithm components maintained in advance to obtain the target algorithm component.

[0074] Step 203: Perform joint training using the target algorithm component.

[0075] The execution subject of this embodiment can be any one of the at least two training devices participating in the joint training. In order to achieve flexible scheduling of multiple algorithms, the corresponding algorithm components are pre-built in this application based on multiple security algorithms, multiple training algorithms, multiple model verification algorithms, etc. Each algorithm component is used to execute the corresponding algorithm, and each training device maintains each algorithm component. It should be pointed out that each algorithm component can be set in the training device or not, for example, each algorithm component is set in the cloud.

[0076] In the distributed training method provided by the embodiment of the present application, when the training device receives the target scheduling information sent by the scheduling platform, the pre-maintained multiple algorithm components are scheduled according to the target scheduling information to obtain the target algorithm component, and the target algorithm component is used for joint training. Wherein, the target scheduling information is generated after the scheduling platform receives the configuration information submitted by the user, based on the initial scheduling information of at least two training devices participating in the joint training included in the configuration information. Thus, by pre-constructing multiple algorithm components and maintaining the multiple algorithm components by each training device, and providing a configuration function through the scheduling platform, the scheduling platform can determine the target scheduling information according to the initial scheduling information input by the user, thereby sending the target scheduling information to the training device to schedule the multiple algorithm components maintained by the training device. It can be seen that the flexible scheduling of multiple algorithms is realized in this training method. In addition, since the training device can directly perform joint training based on the scheduled target algorithm component without the need to perform code writing and other operations based on the corresponding training requirements, the training efficiency is greatly improved.

[0077] In order to facilitate the maintenance of multiple algorithm components, in some implementations, the multiple algorithm components can be divided into multiple component sets, that is, the multiple algorithm components can include multiple security components in the security component set, multiple training components in the training component set, and multiple verification components in the verification component set, each security component is used to execute a security algorithm, each training component is used to execute a training algorithm, and each verification component is used to execute a verification algorithm. Figure 6 As shown, step 202 may include the following steps 2021 and 2022.

[0078] Step 2021, if the security algorithm, training algorithm and model verification algorithm are obtained from the target scheduling information, a security component for executing the security algorithm is selected from the security component set to obtain a target security component; a component for executing training of the training algorithm is selected from the training component set to obtain a target training component; and a verification component for executing the model verification algorithm is selected from the verification component set to obtain a target verification component.

[0079] Specifically, a third association relationship between each security algorithm and the first component identifier of each security component, a fourth association relationship between each training algorithm and the second component identifier of each training component, and a fifth association relationship between each model verification algorithm and the third component identifier of each verification component can be pre-set. Accordingly, when the training device obtains the security algorithm, the training algorithm and the model verification algorithm from the target scheduling information, the associated target first component identifier is queried in the third association relationship according to the obtained security algorithm, and the security component corresponding to the target first component identifier in the security component set is determined as the target security component for executing the security algorithm; the associated target second component identifier is queried in the fourth association relationship according to the obtained training algorithm, and the training component corresponding to the target second component identifier in the training component set is determined as the target training component for executing the training algorithm; and the associated target third component identifier is queried in the fifth association relationship according to the obtained model verification algorithm, and the verification component corresponding to the target third component identifier in the verification component set is determined as the target verification component for executing the model verification algorithm; and the target security component, the target training component and the target verification component are determined as the target algorithm component.

[0080] It is understandable that when the training device obtains the training algorithm from the target scheduling information, it can be determined that the training device is the master device of the joint training, and step 2031 is executed.

[0081] Step 2022, if the security algorithm and the model verification algorithm are obtained from the target scheduling information, a security component for executing the security algorithm is selected from the security component set to obtain a target security component; and a verification component for executing the model verification algorithm is selected from the verification component set to obtain a target verification component.

[0082] Among them, the selection method of the target security component and the target verification component can refer to the relevant description in the previous text, and the repeated parts will not be repeated here.

[0083] It is understandable that when the training device fails to obtain the training algorithm from the target scheduling information, it can be determined that the training device is a slave device of the joint training, and step 2036 is executed.

[0084] Thus, the training device implements the scheduling of each component in the corresponding algorithm component set based on each algorithm obtained from the target scheduling information.

[0085] After the target algorithm component is obtained through scheduling, joint training can be performed based on the target algorithm component. As mentioned above, in distributed training, the master device and the slave device obtain different algorithms from the target scheduling information, and accordingly, the operations performed based on the scheduled target algorithm component are also different. Specifically, Figure 6 As shown, step 203 may include the following steps 2031 to 2039: Step 2031, receiving a desensitized training sample sent by a slave device of a training device.

[0086] Specifically, when the training device determines that the training device is the master device of the joint training according to the target scheduling information, the desensitized training samples sent by the slave device of the training device are received. Among them, determining that the training device is the master device of the joint training according to the target scheduling information may include: if the training device obtains the training algorithm from the target scheduling information, then determining that the training device is the master device of the joint training. Alternatively, if the label information in the sample information in the target scheduling information represents that the first training data of the training device carries a label, then determining that the training device is the master device. The desensitized training samples are obtained by the slave device desensitizing its second training samples. The second training samples can be various types of training samples such as images, texts, audio and video. The operation process of the slave device can be found in the relevant description below, and the repetitions will not be repeated here.

[0087] Step 2032: Acquire a plurality of first training samples of the training device according to the sample information in the target scheduling information.

[0088] Specifically, sample information is obtained from the target scheduling information, and multiple first training samples of the training device are obtained according to the access information in the sample information. The first training samples may be training samples of various types such as images, texts, audio and video.

[0089] It is understandable that the first training sample and the second training sample may be different depending on the target model of the joint training. In a specific embodiment, the target model may be a risk detection model for detecting whether a user has a risk of default, and accordingly, the first training sample and the second training sample may be behavior records in text form, which may include consumption records, repayment records, etc. In another specific embodiment, the target model may be a target detection model for detecting pharmaceutical logos, and accordingly, the first training sample and the second training sample may be image data of pharmaceutical products. In yet another specific embodiment, the target model may be an audio detection model, and accordingly, the first training sample and the second training sample may be audio data. In this regard, they will not be listed one by one in this application, and they can be set as needed in actual applications.

[0090] In some embodiments, the access information in the sample information may be an access address of a database where the first training sample is located; accordingly, the training device obtains the first training sample according to the access address. In some embodiments, the access information in the sample information may also be a library identifier of a database where the first training sample is located; accordingly, the training device obtains the associated access address from the sixth association relationship between the library identifier and the access address according to the library identifier, and obtains the first training sample according to the access address.

[0091] It should be noted that the execution order of step 2031 and step 2032 can be interchanged and can also be executed simultaneously.

[0092] Step 2033, training the first model of the training device using the target training component, the desensitized training sample and the first training sample to obtain a training result and a trained first model; In order to ensure the data security of the training samples, step 2033 may include: using the second security component in the target security component to perform privacy set intersection processing on the first training sample to obtain the target first training sample; and using the target training component, the desensitized training sample and the target first training sample to train the first model of the training device to obtain the training results and the trained first model.

[0093] It should be pointed out that when the training mode is the interactive training mode, the first model of the training device is trained using the target training component, the desensitized training sample, and the target first training sample corresponding to the desensitized training sample. Since the privacy set intersection processing can find the common training samples between the second training sample and the first training sample under the premise of protecting the data privacy of the master device and the slave device, the target first training sample corresponding to the desensitized training sample has a corresponding relationship with the second training sample corresponding to the desensitized training sample. When the training mode is the centralized training mode, in each training round, the first model of the training device is also trained using the desensitized training samples and the target first training samples that have a corresponding relationship.

[0094] Step 2034: Desensitize the training result using the first security component in the target security component to obtain a desensitized training result, and send the desensitized training result to the slave device.

[0095] In different training modes, the training results are processed differently. Figure 7 As shown in the schematic diagram of the interactive training mode, if the training mode in the target scheduling information is the interactive training mode, the first security component is used to perform desensitization processing on the first training result of the current training round to obtain a desensitized training result, and the desensitized training result is sent to the slave device. Figure 8 The schematic diagram of the centralized training mode shown in the figure shows that if the training mode in the target scheduling information is the centralized training mode, the first security component is used to desensitize the second training results obtained after each round of training to obtain desensitized training results, and the desensitized training results are sent to the slave device.

[0096] Among them, the first training result and the second training result may include data for adjusting model parameters, and the data for adjusting model parameters may be at least one of the gradient value, forward reasoning value, etc. of the corresponding training round. The second training result is the training result when the training end condition is met. The training end condition may be that the training round reaches a preset round, or that the loss value determined based on the loss calculation is less than the preset loss value. The training end condition can be set by itself as needed in actual applications. The desensitized training result is the training result after desensitizing the sensitive information in the training result. The sensitive information may be the data used to adjust the model parameters.

[0097] In some embodiments, when the training mode is the interactive training mode, the master device may also send confirmation information to the slave device after receiving each desensitized training sample sent by the slave device. The confirmation information may include the number of desensitized training samples, the number of features in the desensitized training samples, etc., to determine that no data is lost during the transmission of the desensitized data.

[0098] Therefore, in different training modes, different methods are used to process the training results to ensure the effectiveness of the training.

[0099] Step 2035: If it is determined based on the training results that the model verification condition is met, the target verification component is used to verify the first model when the model verification condition is met to obtain a first verification result.

[0100] Among them, the first verification result characterizes the accuracy of the first model. In some embodiments, model verification can be performed after obtaining the target model. Accordingly, if it is determined according to the training results that the model verification conditions are met, it can include: if it is determined that each training sample has participated in the training, and the loss value determined according to the training results is less than the preset loss value, then it is determined that the model verification conditions are met. In some embodiments, model verification can also be performed after each certain round of training. Accordingly, if it is determined according to the training results that the model verification conditions are met, it can include: if it is determined that the training rounds corresponding to the training results reach the preset rounds, then it is determined that the model verification conditions are met.

[0101] It can be understood that when the training device is the master device, after executing step 2021, the above steps 2031 to 2035 are executed.

[0102] Step 2036: Acquire multiple second training samples for the training device according to the sample information in the target scheduling information.

[0103] Specifically, when the training device determines that the training device is a slave device of the joint training according to the target scheduling information, multiple second training samples of the training device are obtained according to the sample information in the target scheduling information. The process of obtaining the second training samples is the same as the process of obtaining the first training samples, and the relevant descriptions in the previous text can be referred to, and the repeated parts will not be repeated here.

[0104] Step 2037: Desensitize the second training sample using the first security component in the target security component to obtain a desensitized training sample, and send the desensitized training sample to the main device.

[0105] In order to ensure the data security of the second training sample, step 2037 may include: using the second security component in the target security component to perform privacy set intersection processing on the second training sample to obtain the target second training sample; and using the first security component in the target security component to desensitize the target second training sample to obtain a desensitized training sample, and sending the desensitized training sample to the main device.

[0106] Furthermore, in order to ensure that the master device and the slave device adopt the same training mode for joint training while ensuring the safety of the samples, the target scheduling information may also include the training mode. Accordingly, step 2037 may include: if the training mode is an interactive training mode, then each second training sample is divided and processed to obtain multiple training sample subsets; the training sample subset corresponding to the current training round is desensitized using the first security component to obtain a desensitized training sample; if the training mode is a centralized training mode, then the first security component is used to desensitize each second training sample to obtain a desensitized training sample.

[0107] Specifically, Figure 7 As shown, if the training mode is the interactive training mode, the second security component in the target security component is used to perform privacy set intersection processing on the second training sample to obtain the target second training sample; the target second training sample is divided to obtain multiple training sample subsets; the first security component is used to perform desensitization processing on the training sample subset corresponding to the current training round to obtain desensitized training samples; and the desensitized training samples are sent to the main device. Figure 8 As shown, if the training mode is a centralized training mode, the second security component in the target security component is used to perform privacy set intersection processing on the second training sample to obtain the target second training sample; and the first security component in the target security component is used to desensitize each target second training sample to obtain a desensitized training sample, and the desensitized training sample is sent to the main device.

[0108] It is understandable that if Figure 7 As shown, when the training mode is the interactive training mode, since the joint training is an iterative multi-round training, in step 1-15, when the slave device determines that the training end condition is not met, it returns to step 1-9 to perform the next round of training; in step 1-17, when the master device determines that the training end condition is not met, it returns to step 1-11 to perform the next round of training. Figure 8 As shown, when the training mode is the centralized training mode, in step 2-11, the master device performs multiple rounds of training on the first model. The first target model obtained by the master device and the second target model obtained by the slave device may be the same or different.

[0109] It should be pointed out that Figure 7 and Figure 8 For illustration only and not for limitation, the execution order of some steps can be interchanged or performed simultaneously. For example, Figure 7 Steps 1-1 to 1-4 and steps 1-5 to 1-9 can be performed simultaneously. Figure 8Steps 2-1 to 2-4 and steps 2-5 to 2-9 can be executed simultaneously, that is, after the master device and the slave device receive the target scheduling information respectively, they can execute the corresponding operations by themselves, and there is no order between them. Figure 7 Steps 1-16 and 1-17 in the method may be performed before step 1-12, or may be performed simultaneously with step 1-12. Figure 8 Steps 2-15 can also be executed at any time during multiple rounds of training. Figure 7 and Figure 8 For the specific implementation process of each step, please refer to the relevant description in the previous article, and the repeated parts will not be repeated here.

[0110] Step 2038, receiving the desensitization processing result sent by the main device.

[0111] Step 2039: If it is determined according to the desensitization processing result that the parameter adjustment conditions are met, the second model of the training device is subjected to parameter adjustment processing according to the desensitization processing result; if it is determined according to the desensitization processing result that the model verification conditions are met, the target verification component is used to verify the second model when the model verification conditions are met to obtain a second verification result.

[0112] Specifically, if model verification is performed after obtaining the target model, then if it is determined in step 2039 that the model verification condition is met according to the desensitization processing result, it may include: if it is determined according to the desensitization processing result that the target model is obtained, then it is determined that the model verification condition is met. If model verification is performed after each training round, then if it is determined in step 2039 that the model verification condition is met according to the desensitization processing result, it may include: if it is determined that the training round corresponding to the desensitization processing result reaches the preset round, then it is determined that the model verification condition is met. The second verification result represents the accuracy of the second model.

[0113] It can be understood that when the training device is a slave device, after executing step 2022, the above steps 2036 to 2039 are executed.

[0114] It can be seen that no matter whether the training device is a master device or a slave device, it can schedule multiple algorithm components according to the security algorithm, model verification algorithm, etc. in the target scheduling information, and perform joint training based on the scheduled target algorithm components. This not only realizes the flexible scheduling of multiple algorithms, but also helps to improve the training rate.

[0115] Corresponding to the aforementioned embodiments of the distributed training method, the present application also provides embodiments of a distributed training device. Fig. 9 is a schematic diagram of a distributed training device according to an exemplary embodiment. The device can be configured in Figure 1 The device is used to execute the distributed training method applied to the scheduling platform provided in any of the above embodiments, such as Fig. 9 As shown, the distributed training device includes: The receiving module 301 is used to receive configuration information submitted by a user, where the configuration information includes initial scheduling information of at least two training devices participating in the joint training; the initial scheduling information includes partial scheduling information of multiple algorithm components.

[0116] The generating module 302 is used to generate target scheduling information of each training device according to the initial scheduling information of each training device; the target scheduling information includes complete scheduling information of multiple algorithm components.

[0117] The sending module 303 is used to send the target scheduling information of each training device to the corresponding training device respectively. The target scheduling information is used to schedule multiple algorithm components based on the target scheduling information, and use the scheduled target algorithm components for joint training. Any training device maintains multiple algorithm components in advance.

[0118] In some implementations, the generation module 302 is specifically configured to: According to the initial scheduling information of each training device, the public configuration information and private configuration information corresponding to each training device are determined; the public configuration information is the same configuration information of at least two training devices, and the private configuration information is the different configuration information of the at least two training devices.

[0119] According to the public configuration information and private configuration information corresponding to each training device, target scheduling information of each training device is generated.

[0120] In some implementations, the generating module 302 is further specifically configured to: The safety algorithm, training algorithm, and model validation algorithm are obtained from the initial scheduling information.

[0121] The training mode is determined according to the security algorithm and the training algorithm; the training mode is an interactive training mode or a centralized training mode.

[0122] The security algorithm, model verification algorithm and training mode are determined as the public configuration information corresponding to each training device.

[0123] In some implementations, the generating module 302 is further specifically configured to: According to the initial scheduling information of each training device, the training algorithm and the communication address corresponding to each training device are determined.

[0124] Determine the sample information of the training samples corresponding to each training device.

[0125] For any training device among the training devices, if the training device is a master device of the joint training, the training algorithm, sample information corresponding to the training device, and the communication address of the slave device of the training device are determined as private configuration information corresponding to the training device.

[0126] If the training device is a slave device of the joint training, the sample information of the training device and the communication address of the master device of the training device are determined as the private configuration information corresponding to the training device.

[0127] In some embodiments, the device further includes: a first acquisition module and a display module.

[0128] The first acquisition module is used to obtain device information of multiple training devices and algorithm information of multiple algorithms to which the user has access rights if it is determined that the login information submitted by the user to the scheduling platform is valid.

[0129] The display module is used to display the configuration interface according to the device information and the algorithm information, and the configuration interface is used to display multiple device information and multiple algorithm information.

[0130] Correspondingly, the receiving module 301 is specifically configured to receive configuration information submitted by the user based on the configuration interface.

[0131] In some embodiments, the device further includes: a determination module.

[0132] The determination module is used to determine the target training device corresponding to the target training sample according to the processing information if the processing information of the target training sample submitted by the user is obtained; the processing information includes the access information of the target training sample.

[0133] The sending module 303 is also used to send a processing request to the target training device according to the processing information; the processing request is used to request the target training device to process the target training sample according to the target structure standard to obtain a standardized training sample; the target structure standard is the structure standard contained in the processing information, or a preset structure standard.

[0134] In some embodiments, the device further includes: a second acquisition module and a processing module.

[0135] The second acquisition module is used to obtain the user's configuration item processing request; the configuration item processing request is used to request new addition processing and / or deletion processing of the target configuration item; the target configuration item includes at least one of a training device, a training sample, a security algorithm, a training algorithm, and a model verification algorithm.

[0136] The processing module is used to process the target configuration item according to the configuration item processing request.

[0137] The distributed training device provided in the embodiment of the present application and the distributed training method provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented therein.

[0138] The implementation process of the functions and effects of each module in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, which will not be repeated here.

[0139] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can refer to the partial description of the method embodiments. The device embodiments described above are merely schematic, wherein the modules described as separate components may or may not be physically separated, and the components illustrated as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application. A person of ordinary skill in the art may understand and implement it without creative work.

[0140] Corresponding to the aforementioned embodiment of the distributed training method, the present application also provides another embodiment of a distributed training device. Fig.10 is a schematic diagram of a distributed training device according to an exemplary embodiment. The device can be configured in Figure 1 The apparatus is used to execute the distributed training method applied to the training device provided in any of the above embodiments, such as Fig.10 As shown, the distributed training device includes: Receiving module 401 is used to receive target scheduling information sent by the scheduling platform; the target scheduling information is generated by the scheduling platform after receiving the configuration information submitted by the user, based on the initial scheduling information of at least two training devices participating in the joint training included in the configuration information; the initial scheduling information includes partial scheduling information of multiple algorithm components, and the target scheduling information includes complete scheduling information of multiple algorithm components.

[0141] The scheduling module 402 is used to schedule the various algorithm components maintained in advance according to the target scheduling information to obtain the target algorithm component.

[0142] The training module 403 is used to perform joint training using the target algorithm component.

[0143] In some implementations, the multiple algorithm components include multiple security components in the security component set, multiple training components in the training component set, and multiple verification components in the verification component set, each security component is used to execute a security algorithm, each training component is used to execute a training algorithm, and each verification component is used to execute a verification algorithm; accordingly, the scheduling module 402 is specifically used to: If the security algorithm, training algorithm and model verification algorithm are obtained from the target scheduling information, a security component for executing the security algorithm is selected from the security component set to obtain the target security component; a training component for executing the training algorithm is selected from the training component set to obtain the target training component; and a verification component for executing the model verification algorithm is selected from the verification component set to obtain the target verification component.

[0144] If the security algorithm and the model verification algorithm are obtained from the target scheduling information, a security component for executing the security algorithm is selected from the security component set to obtain a target security component; and a verification component for executing the model verification algorithm is selected from the verification component set to obtain a target verification component.

[0145] In some implementations, the target scheduling information further includes sample information, the sample information includes access information of the training sample, and the target security component includes the first security component; accordingly, the training module 403 is specifically used to: If it is determined according to the target scheduling information that the training device is the master device of the joint training, then a desensitized training sample sent by a slave device of the training device is received; the desensitized training sample is obtained by the slave device performing desensitization processing on the second training sample.

[0146] A plurality of first training samples of the training device are obtained according to the access information in the sample information.

[0147] The first model of the training device is trained using the target training component, the desensitized training sample and the first training sample to obtain a training result and a trained first model; the training result includes data for adjusting model parameters. The training result is desensitized by using the first security component to obtain a desensitized training result, and the desensitized training result is sent to the slave device; the desensitized training result is a training result after the sensitive information in the training result is desensitized.

[0148] If it is determined based on the training results that the model verification conditions are met, the target verification component is used to verify the first model when the model verification conditions are met to obtain a first verification result; the first verification result represents the accuracy of the first model.

[0149] In some implementations, the target scheduling information further includes a training mode, and the training module 403 is further specifically used to: If the training mode is an interactive training mode, the first security component is used to perform desensitization processing on the first training result of the current training round to obtain a desensitized training result.

[0150] If the training mode is a centralized training mode, the first security component is used to perform desensitization processing on the second training results obtained after each round of training to obtain a desensitized training result.

[0151] In some implementations, the training module 403 is further specifically configured to: If it is determined according to the target scheduling information that the training device is a slave device of the joint training, a plurality of second training samples of the training device are acquired according to the sample information.

[0152] The second training sample is desensitized by using the first security component to obtain a desensitized training sample, and the desensitized training sample is sent to the main device.

[0153] Receive the desensitization training result sent by the main device, where the desensitization training result is the training result after the main device desensitizes the sensitive information in the training result.

[0154] If it is determined based on the desensitization training results that the parameter adjustment conditions are met, the parameters of the second model of the training device are adjusted based on the desensitization training results.

[0155] If it is determined based on the desensitization processing result that the model verification condition is met, the target verification component is used to verify the second model when the model verification condition is met to obtain a second verification result; the second verification result represents the accuracy of the second model.

[0156] In some implementations, the target scheduling information further includes a training mode; accordingly, the training module 403 is further specifically configured to: If the training mode is the interactive training mode, each second training sample is divided and processed to obtain a plurality of training sample subsets; the training sample subset corresponding to the current training round is desensitized using the first security component to obtain a desensitized training sample.

[0157] If the training mode is a centralized training mode, the first security component is used to perform desensitization processing on each second training sample to obtain a desensitized training sample.

[0158] In some implementations, the target security component further includes a second security component; and the training module 403 is further specifically configured to: The first training sample is subjected to privacy set intersection processing by using the second security component to obtain a target first training sample; and the first model of the training device is trained by using the target training component, the desensitized training sample and the target first training sample.

[0159] The second security component is used to perform privacy set intersection processing on the second training sample to obtain a target second training sample; and the first security component is used to perform desensitization processing on the target second training sample to obtain a desensitized training sample.

[0160] The distributed training device provided in the embodiment of the present application and the distributed training method provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented therein.

[0161] The implementation process of the functions and effects of each module in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, which will not be repeated here.

[0162] Corresponding to the aforementioned embodiments of the distributed training method, the present application also provides embodiments of a distributed training system. Fig.11 FIG. 1 is a schematic diagram of a distributed training system according to an exemplary embodiment. Fig.11 As shown, the distributed training system includes: a scheduling platform 501 and multiple training devices 502.

[0163] The configuration scheduling platform 501 is used to receive configuration information submitted by a user, wherein the configuration information includes initial scheduling information of at least two training devices 502 participating in joint training among multiple training devices 502; target scheduling information of at least two training devices 502 is generated based on the initial scheduling information of each of the at least two training devices 502; the target scheduling information of at least two training devices 502 is sent to the corresponding training devices 502 respectively; the initial scheduling information includes partial scheduling information of multiple algorithm components, and the target scheduling information includes complete scheduling information of multiple algorithm components.

[0164] The training device 502 is used to receive the target scheduling information sent by the scheduling platform 501; schedule and process the pre-maintained multiple algorithm components according to the target scheduling information to obtain the target algorithm component; and perform joint training using the target algorithm component.

[0165] The distributed training system provided in the embodiment of the present application and the distributed training method provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented therein.

[0166] It can be understood that the division of the above modules is merely a division of logical functions. In actual implementation, the functions of the above modules can be integrated into hardware entity implementations. For example, the functions of the aforementioned receiving module 301, sending module 303 and receiving module 401 can be integrated into the transceiver implementation, and the functions of the aforementioned generation module 302, scheduling module 402 and training module 403 can be integrated into the processor implementation, etc.

[0167] Based on this, some embodiments of the present application also provide an electronic device corresponding to the distributed training method provided in the aforementioned embodiment, and the electronic device can be configured as follows: Figure 1 The scheduling platform or training device shown is used to execute the above-mentioned distributed training method.

[0168] Fig.12The hardware structure diagram of an electronic device according to an exemplary embodiment is shown, and the electronic device includes: a communication interface 601, a processor 602, a memory 603 and a bus 604; wherein the communication interface 601, the processor 602 and the memory 603 communicate with each other through the bus 604. The processor 602 can execute the distributed training method described above by reading and executing the machine executable instructions corresponding to the control logic of the distributed training method in the memory 603. The specific content of the method is referred to the above embodiment, and will not be repeated here.

[0169] The memory 603 mentioned in the embodiment of the present application can be any electronic, magnetic, optical or other physical storage device, and can contain storage information, such as executable instructions, data, etc. Specifically, the memory 603 can be RAM (Random Access Memory), flash memory, storage drive (such as hard disk drive), any type of storage disk (such as optical disk, DVD, etc.), or similar storage medium, or a combination thereof. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 601 (which can be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. can be used.

[0170] The bus 604 may be an ISA bus, a PCI bus or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory 603 is used to store programs, and the processor 602 executes the programs after receiving execution instructions.

[0171] Processor 602 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in processor 602 or an instruction in software form. The above-mentioned processor 602 can be a general-purpose processor, including a network processor (Network Processor, referred to as NP), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a readily available programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware controls, etc. The disclosed methods, steps and logic block diagrams in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as a hardware decoding processor to execute, or the hardware and software modules in the decoding processor are combined to execute.

[0172] The electronic device provided in the embodiment of the present application and the distributed training method provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented therein.

[0173] The present application also provides a computer-readable storage medium corresponding to the distributed training method provided in the above embodiment. Fig.13 As shown, the computer-readable storage medium shown is a CD 70 on which a computer program (ie, a program product) is stored. When the computer program is run by a processor, the distributed training method provided by any of the aforementioned embodiments will be executed.

[0174] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical or magnetic storage media, which are not listed here one by one.

[0175] The computer-readable storage medium provided in the above-mentioned embodiments of the present application and the distributed training method provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0176] An embodiment of the present application also provides a computer program product corresponding to the distributed training method provided in the aforementioned embodiment. The computer program product includes a computer program, and the computer program is executed by a processor to implement the distributed training method provided in the aforementioned embodiment.

[0177] The computer program product provided by the above-mentioned embodiments of the present application and the distributed training method provided by the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0178] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the claims.

[0179] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0180] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A distributed training method, characterized in that: Applied to a scheduling platform, the method includes: receiving configuration information submitted by a user; the configuration information includes initial scheduling information of at least two training devices participating in the joint training, and the initial scheduling information includes partial scheduling information of multiple algorithm components; Generating target scheduling information of the training device according to the initial scheduling information of the training device; the target scheduling information includes complete scheduling information of the multiple algorithm components; The target scheduling information of the training device is sent to the training device respectively, the target scheduling information is used to schedule the multiple algorithm components based on the target scheduling information, and joint training is performed based on the scheduled target algorithm components. The training device maintains the multiple algorithm components in advance.

2. The method according to claim 1, characterized in that The step of generating target scheduling information of the training device according to the initial scheduling information of the training device comprises: Determine the public configuration information and private configuration information corresponding to the training device according to the initial scheduling information of the training device; the public configuration information is the same configuration information of the at least two training devices, and the private configuration information is the different configuration information of the at least two training devices; The target scheduling information of the training device is generated according to the public configuration information and the private configuration information corresponding to the training device.

3. The method according to claim 2, characterized in that The determining, according to the initial scheduling information of the training device, the public configuration information corresponding to the training device comprises: Obtaining a safety algorithm, a training algorithm, and a model verification algorithm from the initial scheduling information; Determine a training mode according to the safety algorithm and the training algorithm; the training mode includes an interactive training mode or a centralized training mode; The security algorithm, the model verification algorithm and the training mode are determined as public configuration information corresponding to the training device.

4. The method according to claim 2 or 3, characterized in that: Determining private configuration information corresponding to the training device according to the initial scheduling information of the training device includes: Determine, according to the initial scheduling information of the training device, a training algorithm and communication addresses corresponding to the training devices respectively; Determining sample information of training samples respectively corresponding to the training devices; If the training device is a master device of the joint training, determining the training algorithm, sample information corresponding to the training device, and a communication address of a slave device of the training device as private configuration information corresponding to the training device; If the training device is a slave device of the joint training, the sample information of the training device and the communication address of the master device of the training device are determined as the private configuration information corresponding to the training device.

5. The method according to any one of claims 1 to 3, characterized in that: Before receiving the configuration information submitted by the user, the method further includes: If it is determined that the login information submitted by the user to the scheduling platform is valid, obtaining device information of the training device and algorithm information of the algorithm to which the user has access rights; Display a configuration interface according to the device information and the algorithm information, wherein the configuration interface is used to display the device information and the algorithm information; The receiving of configuration information submitted by the user includes: Receive configuration information submitted by the user based on the configuration interface.

6. The method according to any one of claims 1 to 3, characterized in that: The method further comprises: If processing information of the target training sample submitted by the user is obtained, the target training device corresponding to the target training sample is determined according to the processing information; the processing information includes access information of the target training sample; According to the processing information, a processing request is sent to the target training device; the processing request is used to request the target training device to process the target training sample according to the target structure standard to obtain a standardized training sample; the target structure standard is the structure standard contained in the processing information, or a preset structure standard.

7. The method according to any one of claims 1 to 3, characterized in that: The method further comprises: Obtaining a configuration item processing request from a user; the configuration item processing request is used to request adding and / or deleting a target configuration item; the target configuration item includes at least one of a training device, a training sample, a security algorithm, a training algorithm, and a model verification algorithm; According to the configuration item processing request, the target configuration item is processed accordingly.

8. A distributed training method, characterized in that: Applied to a training device, the method comprises: Receive target scheduling information sent by the scheduling platform; the target scheduling information is generated by the scheduling platform after receiving the configuration information submitted by the user, based on the initial scheduling information of at least two training devices participating in the joint training included in the configuration information; the initial scheduling information includes partial scheduling information of multiple algorithm components, and the target scheduling information includes complete scheduling information of the multiple algorithm components; According to the target scheduling information, scheduling processing is performed on the plurality of algorithm components maintained in advance to obtain a target algorithm component; The target algorithm components are used for joint training.

9. The method according to claim 8, characterized in that The multiple algorithm components include multiple security components in the security component set, multiple training components in the training component set, and multiple verification components in the verification component set, each security component is used to execute a security algorithm, each training component is used to execute a training algorithm, and each verification component is used to execute a verification algorithm; The step of performing scheduling processing on the plurality of algorithm components maintained in advance according to the target scheduling information to obtain the target algorithm component includes: If a security algorithm, a training algorithm and a model verification algorithm are obtained from the target scheduling information, a security component for executing the security algorithm is selected from the security component set to obtain a target security component; a training component for executing the training algorithm is selected from the training component set to obtain a target training component; and selecting a verification component for executing the model verification algorithm from the verification component set to obtain a target verification component; If a security algorithm and a model verification algorithm are obtained from the target scheduling information, a component for executing the security algorithm is selected from the security component set to obtain a target security component; and a component for executing the model verification algorithm is selected from the verification component set to obtain a target verification component.

10. The method according to claim 9, characterized in that The target scheduling information further includes sample information, the sample information includes access information of training samples, the target security component includes a first security component; and the joint training using the target algorithm component includes: If it is determined according to the target scheduling information that the training device is the master device of the joint training, receiving a desensitized training sample sent by a slave device of the training device; the desensitized training sample is obtained by the slave device performing desensitization processing on the second training sample; Acquire a plurality of first training samples of the training device according to the access information; Using the target training component, the desensitized training sample and the first training sample, the first model of the training device is trained to obtain a training result and a trained first model; the training result includes data for adjusting model parameters; Desensitizing the training result by using the first security component to obtain a desensitized training result, and sending the desensitized training result to the slave device; the desensitized training result is a training result after desensitizing sensitive information in the training result; If it is determined according to the training result that the model verification condition is met, the target verification component is used to verify the first model when the model verification condition is met to obtain a first verification result; the first verification result represents the accuracy of the first model.

11. The method according to claim 10, characterized in that The target scheduling information also includes a training mode; and the desensitizing the training result by using the first security component to obtain a desensitized training result includes: If the training mode is an interactive training mode, desensitizing the first training result of the current training round by using the first security component to obtain a desensitized training result; If the training mode is a centralized training mode, the first security component is used to perform desensitization processing on the second training results obtained after each round of training to obtain a desensitized training result.

12. The method according to claim 10, characterized in that The joint training using the target algorithm component also includes: If it is determined according to the target scheduling information that the training device is a slave device of the joint training, acquiring a plurality of second training samples of the training device according to the sample information; Using the first security component to perform desensitization processing on the second training sample to obtain a desensitized training sample, and sending the obtained desensitized training sample to a master device of the training device; Receive a desensitization training result sent by the master device; the desensitization training result is a training result after the master device desensitizes sensitive information in the training result; If it is determined that the parameter adjustment condition is met according to the desensitization training result sent by the master device, parameter adjustment processing is performed on the second model of the training device according to the desensitization training result sent by the master device; If it is determined that the model verification conditions are met based on the desensitization training results sent by the main device, the target verification component is used to verify the second model when the model verification conditions are met to obtain a second verification result; the second verification result represents the accuracy of the second model.

13. The method according to claim 12, characterized in that The target scheduling information also includes a training mode, and the desensitizing the second training sample using the first security component to obtain a desensitized training sample includes: If the training mode is an interactive training mode, each of the second training samples is divided and processed to obtain a plurality of training sample subsets; the first security component is used to perform desensitization processing on the training sample subset corresponding to the current training round to obtain a desensitized training sample; If the training mode is a centralized training mode, the first security component is used to perform desensitization processing on each of the second training samples to obtain a desensitized training sample.

14. The method according to claim 12, characterized in that The target security component also includes a second security component; and the training of the first model of the training device using the target training component, the desensitized training sample and the first training sample includes: Using the second security component to perform privacy set intersection processing on the first training sample to obtain a target first training sample; using the target training component, the desensitized training sample and the target first training sample to train the first model of the training device; the target first training sample is a training sample shared by the first training sample and the second training sample of the slave device; The using the first security component to perform desensitization processing on the second training sample to obtain a desensitized training sample includes: The second security component is used to perform the privacy set intersection processing on the second training sample to obtain a target second training sample; and the first security component is used to perform desensitization processing on the target second training sample to obtain a desensitized training sample; the target second training sample is a training sample common to the second training sample and the first training sample of the main device.

15. A distributed training device, characterized in that: Applied to the dispatching platform, the device comprises: A receiving module, configured to receive configuration information submitted by a user; the configuration information includes initial scheduling information of at least two training devices participating in the joint training, and the initial scheduling information includes partial scheduling information of multiple algorithm components; A generating module, configured to generate target scheduling information of the training device according to the initial scheduling information of the training device; the target scheduling information includes complete scheduling information of the multiple algorithm components; A sending module is used to send the target scheduling information of the training device to the training device respectively, the target scheduling information is used to schedule the multiple algorithm components based on the target scheduling information, and perform joint training based on the scheduled target algorithm components, and the training device pre-maintains the multiple algorithm components.

16. A distributed training device, characterized in that: Applied to training equipment, the device comprises: A receiving module, configured to receive target scheduling information sent by a scheduling platform; the target scheduling information is generated by the scheduling platform after receiving the configuration information submitted by the user, based on the initial scheduling information of at least two training devices participating in the joint training included in the configuration information; the initial scheduling information includes partial scheduling information of multiple algorithm components, and the target scheduling information includes complete scheduling information of the multiple algorithm components; A scheduling module, configured to perform scheduling processing on the plurality of algorithm components maintained in advance according to the target scheduling information to obtain a target algorithm component; A training module is used to perform joint training using the target algorithm component.

17. A distributed training system, characterized in that: Includes: a dispatching platform and multiple training devices; The configuration and scheduling platform is used to receive configuration information submitted by a user, wherein the configuration information includes initial scheduling information of at least two training devices participating in the joint training among the multiple training devices; generating target scheduling information of the training device according to the initial scheduling information of the training device; Sending target scheduling information of the training device to the training device respectively; the initial scheduling information includes partial scheduling information of multiple algorithm components, and the target scheduling information includes complete scheduling information of the multiple algorithm components; The training device is used to receive the target scheduling information sent by the scheduling platform; schedule the multiple algorithm components maintained in advance according to the target scheduling information to obtain the target algorithm component; and use the target algorithm component to perform joint training.

18. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes the program to implement the method according to any one of claims 1 to 7, or implements the method according to any one of claims 8 to 14.

19. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the method according to any one of claims 1 to 7, or to implement the method according to any one of claims 8 to 14.

20. A computer program product comprising a computer program, characterized in that The computer program is executed by a processor to implement the method according to any one of claims 1 to 7, or to implement the method according to any one of claims 8 to 14.

Citation Information

Patent Citations

  • Model acquisition method and equipment

    CN112434462A

  • Federated learning client intelligent selection method and system based on deep reinforcement learning

    CN113191484A

  • Model training method and system, computer readable storage medium and electronic equipment

    CN115438734A

  • Training management method for distributed model training and related device

    CN117407713A

  • Building door opening and closing interlocking rainwater barrier mechanism

    KR1020250040178A