Method, device and equipment for establishing model demand side regulation based on federated learning

Abstract

This disclosure provides a method, apparatus, and device for regulating model demanders based on federated learning. The method includes: receiving data uploaded by model demanders in a federated learning architecture and determining a range of the number of model demanders; generating a data regulation task based on the range of the number of model demanders to create multiple model demanders, and generating corresponding configuration information for each model demander; setting corresponding behavior strategies for each model demander based on their task requests; establishing communication channels between the model demanders and other participants when executing the data regulation task, and generating a demander object corresponding to each model demander based on the configuration information and behavior strategies, and executing the data regulation task based on the demander object. This disclosure simplifies the model demander creation process, reduces labor costs, and improves the efficiency of model demander creation in simulation platforms.

Description

Technical Field

[0001] This disclosure relates to the field of federated learning technology, and in particular to a method, apparatus and equipment for regulating demanders based on federated learning to build models. Background Technology

[0002] With the continuous development of computer technology, the application of artificial intelligence is becoming increasingly widespread. Joint learning, which involves collaborating different stakeholders to conduct machine learning, has become a mainstream trend in training AI models. As a novel distributed machine learning framework, joint learning meets the needs of multiple clients for model training while maintaining data security.

[0003] Before a true federated learning ecosystem is established, it is necessary to verify the rationality and effectiveness of the incentive allocation mechanism design, compare the differences in effects brought about by different incentive mechanism options, and identify shortcomings or loopholes in the incentive allocation mechanism design. Therefore, a simulation verification platform for federated learning incentive allocation mechanisms needs to be designed. Existing simulation verification platforms for incentive allocation mechanisms require a significant amount of manpower to establish model demanders based on federated learning and simulation tasks. The process of establishing model demanders is also quite complex, and errors are prone to occur when manually creating model demanders, reducing the efficiency of model demander creation. Summary of the Invention

[0004] In view of this, the present disclosure provides a method, apparatus and equipment for regulating the establishment of model demand side based on joint learning, so as to solve the problems of the existing technology that the process of establishing model demand side is complicated, requires high human resources, and is prone to errors in the creation of model demand side, thus reducing the creation efficiency.

[0005] A first aspect of this disclosure provides a method for regulating model demanders based on federated learning, comprising: in a federated learning architecture, receiving data uploaded by model demanders and determining a range of the number of model demanders; generating a data regulation task based on the range of the number of model demanders to create multiple model demanders, and generating corresponding configuration information for each model demander, wherein the configuration information includes multiple types of attribute information; setting a corresponding behavior strategy for each model demander based on the task request of the model demander, wherein the behavior strategy includes a strategy for setting a model validation set; when executing the data regulation task, establishing a communication channel between the model demanders and other participants, and generating a demander object corresponding to each model demander based on the configuration information and the behavior strategy, and executing the data regulation task based on the demander object.

[0006] A second aspect of this disclosure provides a control device for establishing model demand parties based on federated learning, comprising: a determining module configured to receive data uploaded by model demand parties in a federated learning architecture and determine a range of the number of model demand parties; a creating module configured to generate a data control task based on the range of the number of model demand parties, so as to create multiple model demand parties according to the data control task and generate corresponding configuration information for each model demand party, wherein the configuration information includes multiple types of attribute information; a setting module configured to set a corresponding behavior strategy for each model demand party according to the task request of the model demand party, wherein the behavior strategy includes a setting strategy for the model validation set; and a control module configured to establish a communication channel between the model demand parties and other participants when executing the data control task, and generate a demand party object corresponding to each model demand party according to the configuration information and the behavior strategy, and execute the data control task based on the demand party object.

[0007] A third aspect of this disclosure provides an electronic device, including 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 steps of the above-described method.

[0008] A fourth aspect of this disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0009] The above-described at least one technical solution adopted in the embodiments of this disclosure can achieve the following beneficial effects:

[0010] In a joint learning architecture, this method receives data uploaded by model requesters and determines the range of the number of model requesters. Based on this range, a data regulation task is generated to create multiple model requesters. For each model requester, corresponding configuration information is generated, including various types of attribute information. According to the task requests of the model requesters, corresponding behavioral strategies are set for each model requester, including strategies for setting the model validation set. When executing the data regulation task, a communication channel is established between the model requesters and other participants. Based on the configuration information and behavioral strategies, a requester object corresponding to each model requester is generated, and the data regulation task is executed based on the requester object. This method can automatically create requesters for simulation tasks in a simulation platform, requiring less human intervention, simplifying the requester creation process, reducing the error rate, and improving the efficiency of requester creation. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of a joint learning architecture provided in an embodiment of this disclosure;

[0013] Figure 2 This is a flowchart illustrating the demand-side regulation method based on joint learning for building a model, as provided in this embodiment of the disclosure.

[0014] Figure 3 This is a schematic diagram of the structure of the control device for the demand side based on joint learning model building provided in this embodiment of the disclosure;

[0015] Figure 4 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this disclosure. Detailed Implementation

[0016] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, so as to provide a thorough understanding of the embodiments of this disclosure. However, those skilled in the art will understand that this disclosure may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this disclosure with unnecessary detail.

[0017] Federation learning refers to the comprehensive utilization of multiple AI (Artificial Intelligence) technologies, under the premise of ensuring data security and user privacy, to collaboratively explore the value of data and foster new intelligent business forms and models based on joint modeling. Federation learning has at least the following characteristics:

[0018] (1) Participating nodes control their own data in a weakly centralized joint training mode to ensure data privacy and security in the process of co-creating intelligence.

[0019] (2) In different application scenarios, various model aggregation optimization strategies are established by using screening and / or combination of AI algorithms and privacy-preserving computing to obtain high-level and high-quality models.

[0020] (3) Under the premise of ensuring data security and user privacy, based on multiple model aggregation optimization strategies, obtain methods to improve the performance of the federated learning engine. The performance methods can be improved by solving problems such as parallel computing architecture, information interaction under large-scale cross-domain networks, intelligent perception, and anomaly handling mechanisms.

[0021] (4) Obtain the needs of multiple users in various scenarios, determine the true contribution of each joint participant through a mutual trust mechanism, and allocate incentives accordingly.

[0022] Based on the above approach, an AI technology ecosystem based on collaborative learning can be established, fully leveraging the value of industry data and promoting the implementation of scenarios in vertical fields.

[0023] A joint learning training method and apparatus according to an embodiment of the present disclosure will now be described in detail with reference to the accompanying drawings.

[0024] Figure 1 This is a schematic diagram of a joint learning architecture provided in an embodiment of this disclosure. Figure 1 As shown, the architecture of joint learning may include a server (central node) 101 and participants 102, 103 and 104.

[0025] In the joint learning process, a basic model can be established through server 101, which then sends this model to participants 102, 103, and 104 with whom it has established a communication connection. Alternatively, any participant can establish the basic model and upload it to server 101, which then sends it to other participants with whom it has established a communication connection. Participants 102, 103, and 104 construct models based on the downloaded basic structure and model parameters, train the models using local data, obtain updated model parameters, and encrypt and upload these updated model parameters to server 101. Server 101 aggregates the model parameters sent by participants 102, 103, and 104 to obtain global model parameters, which are then transmitted back to participants 102, 103, and 104. Participants 102, 103, and 104 iterate on their respective models based on the received global model parameters until the models converge, thus achieving model training. During the collaborative learning process, the data uploaded by participants 102, 103, and 104 are model parameters. Local data is not uploaded to server 101, and all participants can share the final model parameters. Therefore, collaborative modeling can be achieved while ensuring data privacy. It should be noted that the number of participants is not limited to the three mentioned above, but can be set as needed. This embodiment of the disclosure does not impose any restrictions on this.

[0026] Figure 2 This is a flowchart illustrating the demand-side regulation method based on joint learning for building a model, as provided in this embodiment of the disclosure. Figure 2 The demand-side regulation method based on joint learning model building can be executed by the server of the joint learning simulation platform. For example... Figure 2 As shown, this demand-side regulation method based on joint learning can specifically include:

[0027] S201, in the joint learning architecture, receives data uploaded by model requesters and determines the range of the number of model requesters;

[0028] S202, Generate data control tasks based on the range of the number of model demanders, so as to create multiple model demanders according to the data control tasks, and generate corresponding configuration information for each model demander, wherein the configuration information contains multiple types of attribute information;

[0029] S203, Based on the task request of the model requester, set a corresponding behavior strategy for each model requester, which includes the strategy for setting the model validation set.

[0030] S204. When performing data regulation tasks, establish a communication channel between the model demander and other participants, and generate a demander object corresponding to each model demander based on the configuration information and behavior strategy, and perform data regulation tasks based on the demander object.

[0031] Specifically, in a real-world federated learning architecture, there is a central node and multiple participants. When a participant provides a model training request to the central node, that participant can be considered a requester. The central node determines the number of requesters based on the training requests uploaded by the requesters, i.e., how many requesters share the same domain and the same training requirements. Before detailing the technical solution of this disclosure, let's briefly explain the application scenario of this embodiment. In the field of federated learning, when clients participate in federated learning, they inevitably consume device resources, including computing resources, communication resources, and energy. Therefore, without sufficient returns, clients may be unwilling to participate or share their models. Therefore, federated learning needs to design an incentive mechanism to compensate for the aforementioned consumption, either numerically or in terms of models. There are two important objectives in designing a federated learning incentive mechanism: first, to evaluate the contribution of each client; and second, to design a reasonable reward to attract and retain more clients.

[0032] Furthermore, in this embodiment, the demand side is a virtual participant in joint learning created through a simulation platform. Since the simulation platform is used to simulate and verify the allocation incentive mechanism of joint learning, the model demand side here can also be considered a virtualized model demand side in the simulation task of joint learning. In practical application scenarios, the model demand side can be considered as the party that needs the model in joint learning. The model demand side may not participate in the joint learning training process, but when the model demand side is both a demander of the model and a contributor of data, it will participate in the joint learning training process.

[0033] Furthermore, the application scenario of this disclosure embodiment can be considered as follows: when the computer program of the simulation platform starts and a new simulation environment begins, a simulation environment object is generated in the memory of the computer program, and based on the simulation environment object, various entities (including the entity corresponding to the model requester) are created sequentially. In practical applications, the construction of each entity is decoupled and does not depend on other entities for construction.

[0034] According to the technical solution provided in this disclosure, in a joint learning architecture, data uploaded by model requesters is received, and the number range of model requesters is determined. A data regulation task is generated based on the number range of model requesters to create multiple model requesters. Corresponding configuration information is generated for each model requester, including various types of attribute information. Based on the task requests of the model requesters, corresponding behavioral strategies are set for each model requester, including strategies for setting the model validation set. When executing the data regulation task, a communication channel is established between the model requesters and other participants. Based on the configuration information and behavioral strategies, a requester object corresponding to each model requester is generated, and the data regulation task is executed based on the requester object. This disclosure can automatically create requesters for simulation tasks in a simulation platform, requiring less human intervention, simplifying the requester creation process, reducing the error rate of requester creation, and improving the efficiency of requester creation.

[0035] In some embodiments, the data control task includes a data simulation task. Before receiving data uploaded by the model demander and determining the range of the number of model demanders, the method further includes: invoking a preset allocation incentive mechanism algorithm based on joint learning; creating a data simulation task corresponding to the allocation incentive mechanism algorithm according to the allocation incentive mechanism algorithm; and generating a simulation environment object corresponding to the data simulation task in the data simulation platform. The data simulation task is used to perform simulation verification of the allocation incentive mechanism algorithm based on joint learning.

[0036] Specifically, the incentive allocation mechanism is used to encourage more participants in the collaborative learning process. After designing the incentive allocation mechanism, its effectiveness needs to be verified. Therefore, executing corresponding simulation tasks through a simulation platform is crucial for verifying the effectiveness of the incentive allocation mechanism and for identifying any shortcomings or loopholes in its design. It should be noted that in the following embodiments, the simulation platform will be referred to as a data simulation platform or a collaborative learning simulation platform.

[0037] In some embodiments, receiving data uploaded by model requesters and determining the range of the number of model requesters includes: in a created data simulation task, receiving data uploaded by model requesters, determining the range of the number of model requesters based on the identification information in the data, so as to randomly generate a number of model requesters within the range.

[0038] Specifically, after the data simulation task (joint learning simulation task) is created in the data simulation platform (i.e., the joint learning simulation platform), a random integer value n ~ [1, MaxN] is generated for the data simulation task in the data simulation platform. Based on this integer value, n model demanders (hereinafter referred to as demanders) are constructed, and then each demander object is created.

[0039] In some embodiments, corresponding configuration information is generated for each model demander, including: randomly extracting different types of attribute information from a pre-defined attribute information library based on the model demander's identification information, binding the extracted different types of attribute information with the model demander's identification information, and combining the different types of attribute information to obtain configuration information; wherein, the configuration information includes the following information: model type, transaction data, model accuracy requirements, and simulation time step.

[0040] Specifically, when creating n model demanders, the configuration information of each demander is first read. The configuration information can be manually preset, and the computer program can use the manually preset values ​​or randomly generate a series of values ​​within the manually preset range.

[0041] Furthermore, the following is a list and explanation of the contents included in the configuration information:

[0042] Model types: linear regression, logistic regression, SVM, xgboost, neural networks, etc.

[0043] Transaction data: such as the funding budget of the model requesters for joint learning;

[0044] Model accuracy requirement: The accuracy requirement of the model requester for the final joint learning model. The accuracy can be a number between [0, 100].

[0045] Simulation time step: such as setting the time interval for each simulation step, for example, performing a simulation operation every few seconds.

[0046] Here, "model type" refers to the type of model required by the client, i.e., the client's requirement for the type of joint learning model. By specifying the algorithm type, the client can obtain the desired model. For example, if the client requires a linear regression model, then the data contributors in the joint learning simulation platform will participate in the training task of the linear regression model, ultimately resulting in a linear regression model.

[0047] In some embodiments, a corresponding behavioral strategy is set for each model requester based on their task request, including: determining the model request information of the model requester based on their task request; and setting a corresponding behavioral strategy for each model requester based on the model request information and the data simulation task. The behavioral strategy includes whether to set a model validation set, and the difficulty of the model validation set is within a preset difficulty range.

[0048] Specifically, the behavioral strategy of the model requester refers to the strategies employed by the model requester in performing model validation during joint learning simulation tasks. For example, in practical applications, the behavioral strategy of the model requester might involve setting a model validation set. The model validation set is used to validate the model's performance after training. The difficulty of the model validation set is within a preset difficulty range; for example, the difficulty of the model validation set could be a number between [0,1].

[0049] In some embodiments, when performing a data regulation task, establishing a communication channel between the model demander and other participants includes: when determining that a data simulation task is to be performed, establishing a communication channel between the model demander and other simulation entities based on the model demander currently created in the data simulation platform, wherein the simulation entities include a central node created based on a joint learning architecture, and the communication channel is used to transmit model information, allocation of funds information, and public information in the simulation environment.

[0050] Specifically, after generating corresponding configuration information for each model requester and constructing the behavioral strategy for each model requester, it is also necessary to construct communication channels between the model requester and other simulation subjects, and to transmit model information, allocate funding information, and other publicly available information in the simulation environment based on these communication channels.

[0051] Furthermore, the other simulation entities used to establish communication channels with the model requester can be either a central node created based on a joint learning architecture, or other data contributors in the simulation task, or simulation entities corresponding to the model requester. In other words, the model requester can communicate not only with the central node, but also with other model requesters.

[0052] In some embodiments, generating a demand side object corresponding to each model demand side based on configuration information and behavior strategy includes: using the pre-generated configuration information and behavior strategy of the model demand side to configure the parameters of the data structure stored in the memory of the data simulation platform so as to adjust the initial configuration information and initialization parameters corresponding to the initial behavior strategy in the data structure to the latest parameters, and using the data structure corresponding to the demand side after parameter adjustment as the demand side object corresponding to the model demand side.

[0053] Specifically, after establishing a communication channel between the model demander and other simulation subjects, a data structure is generated for each demander in the memory of the computer program corresponding to the data simulation platform. The initialization parameters in the data structure are adjusted using the configuration information and behavior strategy generated above. That is, the original parameters in the data structure are modified using the configuration information and behavior strategy parameters generated above to obtain the final demander object.

[0054] In some embodiments, performing data regulation tasks based on demand-side objects includes: adding demand-side objects to the simulation environment object corresponding to the data simulation task, controlling the demand-side objects to start listening to the communication channel through instructions, so that the data demander corresponding to the demand-side objects can exchange information with other simulation subjects, and enabling the data contributors to generate output results corresponding to the simulation time steps according to the simulation time steps and the received external information.

[0055] Specifically, after creating the demand-side object, it is added to the pre-created simulation environment (i.e., the simulation environment object), thus completing the entire demand-side creation process based on the simulation platform. After adding the demand-side object to the simulation environment, instructions are sent to each demand-side to control them to start listening to the communication channel and exchange information with other simulation entities.

[0056] Furthermore, during the execution of the simulation task, according to the preset simulation time step corresponding to the simulation task, each demander generates an output using the received external information within each simulation time step.

[0057] The following are embodiments of the apparatus disclosed herein, which can be used to execute embodiments of the method disclosed herein. For details not disclosed in the apparatus embodiments of this disclosure, please refer to the embodiments of the method disclosed herein.

[0058] Figure 3 This is a schematic diagram of the structure of the control device for the demand side based on joint learning model building provided in this embodiment of the disclosure. Figure 3 As shown, the demand-side control device based on joint learning model building includes:

[0059] The module 301 is configured to receive data uploaded by model requesters and determine the range of the number of model requesters in the joint learning architecture.

[0060] Module 302 is configured to generate data control tasks based on the range of the number of model demanders, so as to create multiple model demanders according to the data control tasks and generate corresponding configuration information for each model demander, wherein the configuration information contains multiple types of attribute information.

[0061] The setting module 303 is configured to set a corresponding behavior strategy for each model requester based on the task request of the model requester. The behavior strategy includes the setting strategy for the model validation set.

[0062] The control module 304 is configured to establish a communication channel between the model demander and other participants when performing data control tasks, and generate a demander object corresponding to each model demander based on the configuration information and behavior strategy, and perform data control tasks based on the demander object.

[0063] In some embodiments, the data control task includes a data simulation task. Figure 3 Before receiving the data uploaded by the model requester and determining the range of the number of model requesters, the determination module 301 calls the preset allocation incentive mechanism algorithm based on joint learning, creates a data simulation task corresponding to the allocation incentive mechanism algorithm, and generates a simulation environment object corresponding to the data simulation task in the data simulation platform. The data simulation task is used to simulate and verify the allocation incentive mechanism algorithm based on joint learning.

[0064] In some embodiments, Figure 3 In the created data simulation task, the determination module 301 receives the data uploaded by the model requester and determines the number range of model requesters based on the identification information in the data, so as to randomly generate several model requesters within the number range.

[0065] In some embodiments, Figure 3 The creation module 302 randomly extracts different types of attribute information from a pre-defined attribute information library based on the identification information of the model demander, and binds the extracted different types of attribute information with the identification information of the model demander. The different types of attribute information are combined to obtain configuration information. The configuration information includes the following information: model type, transaction data, model accuracy requirements, and simulation time step.

[0066] In some embodiments, Figure 3The setting module 303 determines the model requirement information of the model requirement party based on the task request of the model requirement party, and sets corresponding behavior strategies for each model requirement party based on the model requirement information and data simulation task. The behavior strategies include whether to set a model validation set, and the difficulty of the model validation set is within a preset difficulty range.

[0067] In some embodiments, Figure 3 When the control module 304 determines to execute a data simulation task, it establishes a communication channel between the model demander and other simulation entities based on the model demander currently created in the data simulation platform. The simulation entities include a central node created based on a joint learning architecture. The communication channel is used to transmit model information, fund allocation information, and public information in the simulation environment.

[0068] In some embodiments, Figure 3 The control module 304 uses the pre-generated configuration information and behavior strategy of the model demand side to configure the parameters of the data structure stored in the memory of the data simulation platform, so as to adjust the initial configuration information and initialization parameters corresponding to the initial behavior strategy in the data structure to the latest parameters, and use the data structure after parameter adjustment as the demand side object corresponding to the model demand side.

[0069] In some embodiments, Figure 3 The control module 304 adds the demand side object to the simulation environment object corresponding to the data simulation task, and controls the demand side object to start listening to the communication channel through instructions, so that the model demand side corresponding to the demand side object can exchange information with other simulation subjects, and make the model demand side generate the output results corresponding to the simulation time step according to the simulation time step and the received external information.

[0070] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this disclosure.

[0071] Figure 4 This is a schematic diagram of the structure of the electronic device 4 provided in an embodiment of this disclosure. For example... Figure 4 As shown, the electronic device 4 of this embodiment includes a processor 401, a memory 402, and a computer program 403 stored in the memory 402 and executable on the processor 401. When the processor 401 executes the computer program 403, it implements the steps in the various method embodiments described above. Alternatively, when the processor 401 executes the computer program 403, it implements the functions of each module / unit in the various device embodiments described above.

[0072] For example, computer program 403 may be divided into one or more modules / units, which are stored in memory 402 and executed by processor 401 to perform the present disclosure. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 403 in electronic device 4.

[0073] Electronic device 4 can be a desktop computer, laptop, handheld computer, cloud server, or other electronic device. Electronic device 4 may include, but is not limited to, processor 401 and memory 402. Those skilled in the art will understand that... Figure 4 This is merely an example of electronic device 4 and does not constitute a limitation on electronic device 4. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device may also include input / output devices, network access devices, buses, etc.

[0074] Processor 401 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0075] The memory 402 can be an internal storage unit of the electronic device 4, such as a hard disk or RAM. The memory 402 can also be an external storage device of the electronic device 4, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory 402 can include both internal and external storage units of the electronic device 4. The memory 402 is used to store computer programs and other programs and data required by the electronic device. The memory 402 can also be used to temporarily store data that has been output or will be output.

[0076] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0077] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0078] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0079] In the embodiments provided in this disclosure, it should be understood that the disclosed apparatus / computer devices and methods can be implemented in other ways. For example, the apparatus / computer device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. Multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0080] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0081] Furthermore, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0082] If an integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program may include computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. A computer-readable medium may include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in a computer-readable medium may be appropriately added to or subtracted according to the requirements of legislation and patent practice in a jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0083] The above embodiments are only used to illustrate the technical solutions of this disclosure, and are not intended to limit it. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be included within the protection scope of this disclosure.

Claims

1. A demand-side regulation method based on joint learning to build a model, characterized in that, include: In the joint learning architecture, data uploaded by model requesters is received, and the range of the number of model requesters is determined. Data control tasks are generated based on the number range of the model demanders, so as to create multiple model demanders according to the data control tasks, and generate corresponding configuration information for each model demander, wherein the configuration information includes multiple types of attribute information; Based on the task request of the model requester, a corresponding behavior strategy is set for each model requester, and the behavior strategy includes a strategy for setting the model validation set. When the data regulation task is executed, a communication channel is established between the model demander and other participants, and a demander object corresponding to each model demander is generated according to the configuration information and the behavior strategy. The data regulation task is then executed based on the demander object. The data control task includes a data simulation task. Before receiving the data uploaded by the model requester and determining the range of the number of model requesters, the method further includes: A preset allocation incentive mechanism algorithm based on joint learning is invoked. According to the allocation incentive mechanism algorithm, a data simulation task corresponding to the allocation incentive mechanism algorithm is created, and a simulation environment object corresponding to the data simulation task is generated in the data simulation platform. The data simulation task is used to simulate and verify the allocation incentive mechanism algorithm based on joint learning.

2. The method according to claim 1, characterized in that, The process of receiving data uploaded by model requesters and determining the range of the number of model requesters includes: In the created data simulation task, the data uploaded by the model requester is received, and the number range of the model requesters is determined according to the identification information in the data, so as to randomly generate several model requesters within the number range.

3. The method according to claim 1, characterized in that, The step of generating corresponding configuration information for each model demander includes: Based on the identification information of the model demander, different types of attribute information are randomly extracted from a pre-set attribute information library, and the extracted different types of attribute information are bound with the identification information of the model demander. The different types of attribute information are then combined to obtain the configuration information. The configuration information includes the following: model type, transaction data, model accuracy requirements, and simulation time step.

4. The method according to claim 1, characterized in that, The step of setting corresponding behavioral strategies for each model requester based on their task requests includes: Based on the task request from the model requester, the model requirement information of the model requester is determined. Based on the model requirement information and the data simulation task, a corresponding behavioral strategy is set for each model requester. The behavioral strategy includes whether to set a model validation set, and the difficulty of the model validation set is within a preset difficulty range.

5. The method according to claim 1, characterized in that, When executing the data regulation task, establishing a communication channel between the model demander and other participants includes: When it is determined to execute the data simulation task, a communication channel is established between the model demander and other simulation entities based on the model demander currently created in the data simulation platform. The simulation entities include a central node created based on a joint learning architecture. The communication channel is used to transmit model information, allocation of funds information, and public information in the simulation environment.

6. The method according to claim 1, characterized in that, The step of generating a demand side object corresponding to each of the model demand sides based on the configuration information and the behavior strategy includes: Using the pre-generated configuration information and behavior strategy of the model demander, the parameters of the data structure stored in the memory of the data simulation platform are configured so that the initial configuration information and initialization parameters corresponding to the initial behavior strategy in the data structure are adjusted to the latest parameters, and the data structure with adjusted parameters is used as the demander object corresponding to the model demander.

7. The method according to claim 1, characterized in that, The execution of the data regulation task based on the demand-side object includes: The demand-side object is added to the simulation environment object corresponding to the data simulation task, and the demand-side object is controlled by instructions to start listening to the communication channel so that the model demand-side corresponding to the demand-side object can exchange information with other simulation subjects, and the model demand-side generates the output result corresponding to the simulation time step according to the simulation time step and the received external information.

8. A control device for demand-side modeling based on joint learning, characterized in that, include: The determination module is configured to receive data uploaded by model requesters in the joint learning architecture and determine the range of the number of model requesters. The creation module is configured to generate a data control task based on the number range of the model demanders, so as to create multiple model demanders according to the data control task, and generate corresponding configuration information for each model demander, wherein the configuration information includes multiple types of attribute information. The setting module is configured to set a corresponding behavior strategy for each model requester based on the task request of the model requester, and the behavior strategy includes a setting strategy for the model validation set. The control module is configured to establish a communication channel between the model demander and other participants when executing the data control task, and generate a demander object corresponding to each model demander according to the configuration information and the behavior strategy, and execute the data control task based on the demander object. The data control task includes a data simulation task. The determining module is used to retrieve a preset allocation incentive mechanism algorithm based on joint learning before receiving the data uploaded by the model requester and determining the number range of the model requesters. Based on the allocation incentive mechanism algorithm, a data simulation task corresponding to the allocation incentive mechanism algorithm is created, and a simulation environment object corresponding to the data simulation task is generated in the data simulation platform. The data simulation task is used to perform simulation verification of the allocation incentive mechanism algorithm based on joint learning.

9. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 7.

Images