CN111538598A - Federated learning modeling method, apparatus, device and readable storage medium - Google Patents

Abstract

The present application discloses a federated learning modeling method, apparatus, device, and readable storage medium. The federated learning modeling method includes: negotiating and interacting with each second device associated with the first device, and determining each to-be-completed learning method. model training tasks, and determine each model training participating device corresponding to each of the to-be-completed model training tasks in each of the second devices, and then obtain the model training time period corresponding to each of the to-be-completed model training tasks, and based on For each model training time period, coordinate each of the model training participating devices corresponding to each of the to-be-completed model training tasks to perform a preset federated learning modeling process, so as to complete each of the to-be-completed model training tasks. This application solves the technical problem of the low utilization rate of the coordinator's computing resources in the federated learning system.

Description

Federated learning modeling method, apparatus, device and readable storage medium

technical field

The present application relates to the field of artificial intelligence in financial technology (Fintech), and in particular, to a federated learning modeling method, apparatus, device, and readable storage medium.

Background technique

With the continuous development of financial technology, especially Internet technology finance, more and more technologies (such as distributed, blockchain, artificial intelligence, etc.) are applied in the financial field, but the financial industry also puts forward higher requirements for technology. Requirements, such as the distribution of corresponding to-do items in the financial industry, also have higher requirements.

With the continuous development of computer software and artificial intelligence, the application fields of federated learning are becoming more and more extensive. In a federated learning scenario, a model is usually trained by multiple federated learning participants, and the coordinator is used to coordinate each federated learning. Participants perform model training. For example, in each round of federation, the gradients sent by each federation participant are weighted and averaged, etc. However, when each federation participant performs local iterative training, the coordinator does not need to perform computing tasks but takes up computing power. resources, that is, when each federated participant performs local iterative training, the computing resources of the coordinator are wasted, thereby reducing the utilization of the computing resources of the coordinator. Technical issues with low resource utilization.

SUMMARY OF THE INVENTION

The main purpose of this application is to provide a federated learning modeling method, apparatus, device and readable storage medium, which aims to solve the technical problem of the low utilization rate of the coordinator's computing resources in the federated learning system in the prior art.

To achieve the above purpose, the present application provides a federated learning modeling method, the federated learning modeling method is applied to the first device, and the federated learning modeling method includes:

Negotiate and interact with each second device associated with the first device, determine each model training task to be completed, and determine each model training participating device corresponding to each of the to-be-completed model training tasks in each of the second devices ;

Obtain the model training time periods corresponding to the model training tasks to be completed, and coordinate the model training participating devices corresponding to the model training tasks to be completed to perform preset federated learning based on the model training time periods modeling process to complete each of the to-be-completed model training tasks.

Optionally, the step of determining each model training participating device corresponding to each of the to-be-completed model training tasks in each of the second devices includes:

obtaining model training information corresponding to each of the model training tasks;

Based on each of the model training information, each of the model training participating devices corresponding to each of the to-be-completed model training tasks is determined through a willingness confirmation interaction with each of the second devices.

Optionally, the model training information includes model index information,

The step of determining each of the model training participating devices corresponding to each of the to-be-completed model training tasks by performing a willingness confirmation interaction with each of the second devices based on each of the model training information includes:

Send each of the model index information to each of the second devices, respectively, for each of the second devices to determine in each of the model training tasks based on the acquired model training requirement information and each of the model index information, respectively. Participating in each target model training task, and generating first determination information corresponding to each of the target model training tasks;

Based on each first determination information fed back by each of the second devices, each of the model training participating devices corresponding to each of the to-be-completed model training tasks is determined.

Optionally, the model training information includes model training time information,

The step of determining each of the model training participating devices corresponding to each of the to-be-completed model training tasks by performing a willingness confirmation interaction with each of the second devices based on each of the model training information includes:

Send each of the model training time information to each of the second devices, so that each of the second devices can perform the training tasks in each of the model training tasks based on the acquired training time limit information and the training time information of each of the models. Determine each target model training task to participate in, and generate the second determination information corresponding to each of the target model training tasks;

Based on each second determination information fed back by each of the second devices, each of the model training participating devices corresponding to each of the to-be-completed model training tasks is determined.

Optionally, within each model training time period, receive local model parameters sent by each of the model training participating devices corresponding to the model training time period respectively, and calculate the latest federated model parameters based on a preset aggregation rule;

determining whether the latest federated model parameters meet the preset training task end condition;

If the latest federated model parameters satisfy the preset training task ending condition, sending the latest federated model parameters to each of the second devices, so that each of the second devices can update their respective local models;

If the latest federated model parameters do not meet the preset training task end condition, the latest federated model parameters are sent to each of the model training participating devices respectively, so that each of the model participating devices can update their respective federated participation model to recalculate the latest federated model parameters until the latest federated model parameters satisfy the preset training task end condition.

To achieve the above purpose, the present application also provides a federated learning modeling method, the federated learning modeling method is applied to the second device, and the federated learning modeling method includes:

Interacting with the first device, determining model training information, and acquiring device status information, to determine whether to participate in the to-be-completed model training task corresponding to the model training information based on the device status information;

If participating in the to-be-completed model training task, a preset federated learning modeling process is executed by coordinating and interacting with the first device to complete the to-be-completed model training task.

Optionally, the step of executing a preset federated learning modeling process by coordinating and interacting with the first device includes:

determining the to-be-trained model corresponding to the to-be-completed model training task, and performing iterative training on the to-be-trained model until the to-be-trained model reaches a preset number of iterations, and acquiring local model parameters corresponding to the to-be-trained model;

sending the local model parameters to the first device for the first device to calculate the latest federated model parameters based on the local model parameters;

Receive the latest federated model parameters fed back by the first device, and update the to-be-trained model based on the latest federated model parameters until the local model reaches a preset training end condition, and obtain the corresponding model training tasks to be completed. target modeling model.

The present application also provides a federated learning modeling device, the federated learning modeling device is a virtual device, and the federated learning modeling device is applied to the first device, and the federated learning modeling device includes:

A negotiation module, configured to negotiate and interact with each second device associated with the first device, determine each model training task to be completed, and determine in each of the second devices the corresponding model training tasks to be completed respectively Participating equipment for each model training;

A coordination module, configured to obtain the model training time periods corresponding to the model training tasks to be completed, and coordinate the model training participating devices corresponding to the model training tasks to be completed based on the model training time periods A preset federated learning modeling process is performed to complete each of the to-be-completed model training tasks.

Optionally, the negotiation module includes:

an obtaining unit, configured to obtain model training information corresponding to each of the model training tasks;

A determination unit, configured to determine each of the model training participating devices corresponding to each of the to-be-completed model training tasks, based on each of the model training information, by performing a willingness confirmation interaction with each of the second devices.

Optionally, the determining unit includes:

A first sending subunit, configured to send each of the model index information to each of the second devices, so that each of the second devices can, based on the acquired model training requirement information and each of the model index information, respectively, in the Determining each target model training task involved in each of the model training tasks, and generating first determination information corresponding to each of the target model training tasks;

The first determination subunit is configured to determine each of the model training participating devices corresponding to each of the to-be-completed model training tasks, based on each of the first determination information respectively fed back by each of the second devices.

Optionally, the determining unit further includes:

The second sending subunit is configured to send each of the model training time information to each of the second devices respectively, so that each of the second devices can use the acquired training time limit information and each of the model training time information respectively. , determine each target model training task involved in each of the model training tasks, and generate second determination information corresponding to each of the target model training tasks;

A second determination subunit, configured to determine each of the model training participating devices corresponding to each of the to-be-completed model training tasks based on each of the second determination information fed back by each of the second devices.

Optionally, the coordination module includes:

a computing unit, configured to respectively receive local model parameters sent by each of the model training participating devices corresponding to the model training time period in each of the model training time periods, and calculate the latest federated model parameters based on a preset aggregation rule ;

a first determination unit, configured to determine whether the latest federated model parameters meet the preset training task end condition;

An update unit, configured to send the latest federated model parameters to each of the second devices if the latest federated model parameters satisfy the preset training task end condition, so that each of the second devices can update their respective local model;

A second determination unit, configured to send the latest federated model parameters to each of the model training participating devices, respectively, if the latest federated model parameters do not meet the preset training task ending condition, for each of the The participating devices update their respective federated participation models to recalculate the latest federated model parameters until the newest federated model parameters satisfy the preset training task end condition.

To achieve the above purpose, the present application further provides a federated learning modeling device, the federated learning modeling device is applied to the second device, and the federated learning modeling device further includes:

an interaction module, configured to interact with the first device, determine model training information, and obtain device status information, so as to determine whether to participate in the to-be-completed model training task corresponding to the model training information based on the device status information;

The federated learning modeling module is configured to perform a preset federated learning modeling process by coordinating and interacting with the first device if participating in the to-be-completed model training task to complete the to-be-completed model training task.

Optionally, the federated learning modeling module includes:

an iterative training unit, configured to determine the to-be-trained model corresponding to the to-be-completed model training task, perform iterative training on the to-be-trained model until the to-be-trained model reaches a preset number of iterations, and obtain the corresponding to-be-trained model The local model parameters of ;

a sending unit, configured to send the local model parameters to the first device, so that the first device can calculate the latest federated model parameters based on the local model parameters;

an update unit, configured to receive the latest federated model parameters fed back by the first device, and based on the latest federated model parameters, update the to-be-trained model until the local model reaches a preset training end condition, and obtain the to-be-trained model Complete the target modeling model corresponding to the model training task.

The present application also provides a federated learning modeling device, the federated learning modeling device is an entity device, and the federated learning modeling device includes: a memory, a processor, and a device stored on the memory and available on the processor The program of the federated learning modeling method that runs on the federated learning modeling method, when the program of the federated learning modeling method is executed by the processor, can implement the steps of the federated learning modeling method as described above.

The present application also provides a readable storage medium on which a program for implementing the federated learning modeling method is stored, and when the program for the federated learning modeling method is executed by a processor, the above-mentioned federated learning modeling method is implemented. steps of the modulo method.

The present application determines each model training task to be completed through negotiation and interaction with each second device associated with the first device, and determines each model corresponding to each of the to-be-completed model training tasks in each of the second devices. Training participating equipment, and then obtaining the model training time period corresponding to each model training task to be completed, and based on each model training time period, coordinating each of the model training participating equipment corresponding to each of the model training tasks to be completed. A preset federated learning modeling process is performed to complete each of the to-be-completed model training tasks. That is, the present application provides a method for federated learning based on time division, that is, before performing federated learning modeling, by interacting with each of the second devices, it is determined that each to-be-completed model training needs to be performed task, and then determine each model training participating device and model training time period corresponding to each model training task to be completed, and then the coordinator can coordinate each of the model training tasks to be completed based on each model training time period. Each corresponding model training participating device performs a preset federated learning modeling process to complete each of the to-be-completed model training tasks, that is, each model participating device of the to-be-completed training model is performing local iterative training At the same time, the coordinator can coordinate the model training participating devices corresponding to other model training tasks to be completed to perform federated learning modeling, thereby avoiding the need for the coordinator to perform computing tasks but occupy computing resources when each federated participant performs local iterative training. The situation occurs, and then the purpose of making full use of the coordinator's computing resources is achieved, and the utilization rate of the coordinator's computing resources is improved. Therefore, the technical problem of the low utilization rate of the coordinator's computing resources in the federated learning system is solved.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description serve to explain the principles of the application.

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. In other words, on the premise of no creative labor, other drawings can also be obtained from these drawings.

FIG. 1 is a schematic flowchart of the first embodiment of the federated learning modeling method of the present application;

FIG. 2 is a schematic flowchart of the second embodiment of the federated learning modeling method of the present application;

FIG. 3 is a schematic diagram of the device structure of the hardware operating environment involved in the solution of the embodiment of the present application.

The realization, functional characteristics and advantages of the purpose of the present application will be further described with reference to the accompanying drawings in conjunction with the embodiments.

Detailed ways

It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

An embodiment of the present application provides a federated learning modeling method. In the first embodiment of the federated learning modeling method of the present application, referring to FIG. 1 , the federated learning modeling method is applied to a first device, and the federated learning modeling method is applied to a first device. Modular methods include:

Step S10, negotiate and interact with each second device associated with the first device, determine each model training task to be completed, and determine each model corresponding to each of the to-be-completed model training tasks in each of the second devices training participation equipment;

In this embodiment, it should be noted that one of the model training tasks to be completed corresponds to one or more model training participating devices, the first device is the coordinator of the horizontal federated learning, and the second device is the horizontal federation Learning participants, the model training participating device is the second device participating in the to-be-completed training task, and the to-be-completed model training task is a model training task based on horizontal federated learning, wherein one to-be-completed model training task It can be used to train one or more target models, and one of the target models can also be obtained based on the execution of one or more model training tasks to be completed.

Optionally, the second device may choose to execute the to-be-completed model training task in a preset trusted execution environment, such as Intel's SGX (Intel Software Guard Extensions).

Negotiate and interact with each second device associated with the first device, determine each model training task to be completed, and determine each model training participating device corresponding to each of the to-be-completed model training tasks in each of the second devices Specifically, each second device associated with the first device performs negotiation and interaction to determine each model training task to be completed and the model training information of each model training task to be completed, and then based on the model training information, In each of the second devices, each model training participating device corresponding to each of the to-be-completed model training tasks is determined.

Wherein, in step S10, the step of determining each model training participating device corresponding to each of the to-be-completed model training tasks in each of the second devices includes:

Step S11, acquiring model training information corresponding to each of the model training tasks;

In this embodiment, it should be noted that the model training information includes model name information, model training time period, etc., wherein the model name information is the identifier of the corresponding model to be trained, such as code, character string, etc. , the model training time period is the estimated time information required for model training.

Step S12 , based on each of the model training information, through the willingness confirmation interaction with each of the second devices, determine each of the model training participating devices corresponding to each of the to-be-completed model training tasks respectively.

In this embodiment, based on each of the model training information, each of the model training participating devices corresponding to each of the to-be-completed model training tasks is determined by performing a willingness confirmation interaction with each of the second devices. Specifically, Send each of the model training information to each of the second devices respectively, so that each of the second devices can obtain the device status information, and based on the device status information, respectively determine whether to participate in the pending completion corresponding to each of the model training information The model training task, if it is determined to participate in the model training task to be completed, the determination information corresponding to the model training task to be completed is fed back to the first device, and then the first device receives the determination information respectively, and sends each determination information to the first device. A second device corresponding to the determination information is identified as a model training participating device, and one or more model training participating devices corresponding to each of the to-be-completed model training tasks are counted.

Wherein, the model training information includes model index information,

Described based on each described model training information, by carrying out willingness confirmation interaction with each described second equipment, the step of determining each described model training participating equipment corresponding to each described model training task to be completed respectively comprises:

Step A10: Send each of the model index information to each of the second devices, so that each of the second devices can perform training in each of the models based on the acquired model training requirement information and each of the model index information. Determine each target model training task involved in the task, and generate first determination information corresponding to each of the target model training tasks;

In this embodiment, it should be noted that the model index information is the identification information of the corresponding model training task to be completed, such as a code or a character string, and the first determination information indicates that the second device determines Participate in the information of the to-be-completed model training task corresponding to the model index information, wherein the first determination information may be the willingness information, local model parameter information or local model gradient information that the second device replies alone, to indicate The second device is willing to participate in the corresponding to-be-completed model training tasks, and each of the to-be-completed training tasks corresponds to a model training time period in which the task is performed.

Send each of the model index information to each of the second devices, respectively, for each of the second devices to determine in each of the model training tasks based on the acquired model training requirement information and each of the model index information, respectively. Participate in each target model training task, and generate first determination information corresponding to each target model training task, specifically, broadcast to each of the second devices within a preset time period before the start of each of the model training time periods The model index information corresponding to the model training time period, for the second device to determine the corresponding model training task to be completed based on the model index information, and based on the obtained current device operating state, wherein the device operating state Including currently available computing resources, and then determine whether to participate in the to-be-completed model training task, and if it is determined to participate in the to-be-completed model training task, then feed back first determination information to the first device to indicate participation in the to-be-completed model training task For the training task, if it is determined not to participate in the to-be-completed model training task, the model index information is ignored, and the next model index information is waited for.

Step A20: Determine each of the model training participating devices corresponding to each of the to-be-completed model training tasks, based on each of the first determination information respectively fed back by each of the second devices.

In this embodiment, each of the model training participating devices corresponding to each of the to-be-completed model training tasks is determined based on each first determination information fed back by each of the second devices. Before the training period starts, each of the first determination information corresponding to the to-be-completed model training task corresponding to the model index information sent by each of the second devices is respectively received, and each of the first determination information is sent. Each second device of the information is used as the model training participating device, wherein one of the first determination information corresponds to one of the second devices corresponds to one of the model training participating devices.

Wherein, the model training information includes model training time information,

The step of determining each of the model training participating devices corresponding to each of the to-be-completed model training tasks by performing a willingness confirmation interaction with each of the second devices based on each of the model training information includes:

Step B10: Send each of the model training time information to each of the second devices, so that each of the second devices can use the obtained training time limit information and each of the model training time information in each of the second devices. Determine each target model training task involved in the model training task, and generate second determination information corresponding to each of the target model training tasks;

In this embodiment, it should be noted that the second determination information is information indicating that the second device determines to participate in the to-be-completed model training task corresponding to the model training time information, and each of the to-be-completed training tasks is A model training period corresponding to an execution task, and the second determination information will be sent by the second device to the second device before the model training period corresponding to the model training task to be completed corresponding to the second determination information starts. a device.

Send each of the model training time information to each of the second devices, so that each of the second devices can perform the training tasks in each of the model training tasks based on the acquired training time limit information and the training time information of each of the models. Determine each target model training task to participate in, and generate second determination information corresponding to each target model training task. Specifically, before the model training time period corresponding to each model training task to be completed, each The model training time information is respectively sent to each of the second devices, so that each of the second devices can obtain training time limit information, wherein the training time limit information indicates that the second device is in the model training period. Whether there is free time and enough computing resources to participate in the model training task to be completed within the time period, the second device will determine whether to participate in the model training based on the training restriction information and the model training time information. The model training task to be completed corresponding to the time information, if it is determined to participate in the to-be-completed model training task, the second determination information is fed back to the first device to indicate participation in the to-be-completed model training task, if it is determined not to participate in any model training task. If the model training task to be completed is described, the model training time information is ignored, and the next model training time information is waited for.

Step B20: Determine each of the model training participating devices corresponding to each of the to-be-completed model training tasks based on each of the second determination information fed back by each of the second devices.

In this embodiment, each of the model training participating devices corresponding to each of the to-be-completed model training tasks is determined based on each second determination information fed back by each of the second devices. Before the training period starts, each of the second determination information corresponding to the to-be-completed model training task corresponding to the model training time information sent by each of the second devices is respectively received, and each of the second determination information will be sent. The second device of the second determination information is used as the model training participating device, wherein one of the first determination information corresponds to one of the second device corresponds to one of the model training participating devices.

Step S20: Acquire the model training time period corresponding to each of the model training tasks to be completed, and coordinate the model training participating devices corresponding to the model training tasks to be completed based on the model training time periods to perform pre-preparation. Set up a federated learning modeling process to complete each of the to-be-completed model training tasks.

In this embodiment, it should be noted that the preset federated learning modeling process is a process of performing federated learning, and each of the model training time periods includes a first model training time period and a second model training time period.

Obtain the model training time periods corresponding to the model training tasks to be completed, and coordinate the model training participating devices corresponding to the model training tasks to be completed to perform preset federated learning based on the model training time periods The modeling process is to complete each of the to-be-completed model training tasks, specifically, to obtain the model training time period corresponding to each of the to-be-completed model training tasks, and within each model training time period, receive the corresponding The model trains the local model parameters sent by participating devices, and calculates the latest federated model parameters corresponding to each of the local model parameters based on preset aggregation rules, wherein the preset aggregation rules include weighted average, summation, etc., and determine whether the latest federated model parameters meet the preset training task completion conditions, and if the latest federated model parameters meet the training task completion conditions, send the latest federated model parameters to each of the second devices respectively, for each of the second devices to update their respective local models based on the latest federated model parameters, and if the latest federated model parameters do not meet the training task completion condition, send the latest federated model parameters to each of the models respectively Train participating devices, so that each of the model training participating devices can update their respective local models, and based on the updated local models, perform federated learning again, and recalculate the latest federated model parameters until the latest federated model parameters meet the training task completion conditions , wherein the training task completion conditions include loss function convergence, the model reaches the maximum number of iterations, etc., wherein, if there is an intersection time period for each of the model training time periods, then within the intersection time period, the first device The order of calculating the latest federated model parameters corresponding to each of the to-be-completed model training tasks will be determined according to the chronological order of receiving the local model parameters corresponding to each of the to-be-completed model training tasks. Completing the model training task includes task A and task B, then the first device has all received the local model parameters sent by the model training participating devices corresponding to the task A at 9:07, and at 9:09 , having received all the local model parameters sent by the model training participating devices corresponding to task B, the first device preferentially calculates the latest federated model parameters corresponding to task A, and then calculates the latest federated model parameters corresponding to task B.

Optionally, the first device may choose to execute the step of calculating the latest federated model parameters corresponding to each of the local model parameters based on a preset aggregation rule in the preset trusted execution environment.

Wherein, the step of coordinating each model training participating device corresponding to each model training task to be completed to perform a preset federated learning modeling process based on each model training time period includes:

Step S21, in each of the model training time periods, respectively receiving local model parameters sent by each of the model training participating devices corresponding to the model training time period, and calculating the latest federated model parameters based on a preset aggregation rule;

In this embodiment, it should be noted that the local model parameters include model network parameters, gradient information, etc., wherein the model network parameters are preset by the model training participating device for the local model iterative training held by itself After the number of times, iteratively train the network parameters of the local model. For example, assuming that the local model is a linear model Y=β ₀ +β ₁ X ₁ +β ₂ X ₂ +…+β _n X _n , then the network The parameters are vectors (β ₀ , β ₁ , β ₂ , . . . , β _n ).

During each model training time period, the local model parameters sent by each model training participating device corresponding to the model training time period are respectively received, and the latest federated model parameters are calculated based on the preset aggregation rules. 1. During the model training time period, receive local model parameters sent by each of the model training participating devices corresponding to the model training time period, wherein each of the local model parameters is the result of the model training participating device. The federated participation model corresponding to the local model parameters is obtained by performing iterative training for a preset number of times, wherein the federated participation model is the local model of the model training participating device, and then based on the preset aggregation rules, each local model is The parameters are weighted and averaged to obtain the latest federated model parameters.

Step S22, determining whether the latest federated model parameters meet the preset training task end condition;

In this embodiment, it should be noted that the preset training task ending conditions include training reaching the maximum number of iterations, loss function training convergence, and the like.

Determine whether the latest federated model parameters meet the preset training task end condition, specifically, if the difference between the latest federated model parameters and the last round of the latest federated model parameters is less than a preset difference threshold, determine that the the latest federated model parameters reach the preset training task end condition,

Step S23, if the latest federated model parameters satisfy the preset training task ending condition, send the latest federated model parameters to each of the second devices, so that each of the second devices can update their respective local models ;

In this embodiment, if the latest federated model parameters satisfy the preset training task ending condition, the latest federated model parameters are sent to each of the second devices, so that each of the second devices can update their respective Specifically, if the latest federated model parameters satisfy the preset training task end condition, the latest federated model parameters are sent to each of the second devices for each second device to use based on For the latest federated model parameters, the corresponding model parameters in the local model are replaced and updated to the latest federated model parameters.

In addition, it can also be set that if the difference between the latest federated model parameters and the last round of the latest federated model parameters is smaller than a preset difference threshold for a preset number of times, it is determined that the latest federated model parameters reach The preset training task ending condition.

Step S23, if the latest federated model parameters do not meet the preset training task end condition, send the latest federated model parameters to each of the model training participating devices, so that each of the model participating devices can update their respective models. to recalculate the latest federated model parameters until the latest federated model parameters satisfy the preset training task end condition.

In this embodiment, if the latest federated model parameters do not meet the preset training task ending condition, the latest federated model parameters are sent to each of the model training participating devices respectively for each of the models to participate in. The devices update their respective federated participation models to recalculate the latest federated model parameters until the latest federated model parameters meet the preset training task end condition, specifically, if the latest federated model parameters do not meet the predetermined Set the training task end condition, then send the latest federated model parameters to each of the model training participating devices, so that each of the model training participating devices can update their respective federation participation based on the latest federated model parameters model, and iteratively trains the updated federation participation model, and then when the number of iterative training reaches the preset number of iterative training times, re-acquires the local model parameters of the federated participation model after iterative training, and recalculates each The local model parameters are sent to the first device for the first device to recalculate the latest federation based on the recalculated local model parameters and the preset aggregation rules sent by the second devices model parameters until the latest federated model parameters satisfy the preset training task end condition.

In this embodiment, through negotiation and interaction with each second device associated with the first device, each model training task to be completed is determined, and each of the model training tasks to be completed is determined in each of the second devices. Model training participation equipment, and then obtains the model training time periods corresponding to the model training tasks to be completed, and coordinates the model training participation corresponding to the model training tasks to be completed based on the model training time periods. The device performs a preset federated learning modeling process to complete each of the to-be-completed model training tasks. That is, this embodiment provides a time-division-based method for federated learning, that is, before performing federated learning modeling, each to-be-completed model that needs to be executed is determined by interacting with each of the second devices Training tasks, and then determine each model training participating device and model training time period corresponding to each of the model training tasks to be completed, and then the coordinator can coordinate each of the to-be-completed model training based on each of the model training time periods. Each model training participating device corresponding to the task performs a preset federated learning modeling process to complete each of the to-be-completed model training tasks, that is, each model participating device of the to-be-completed training model is performing local iteration During training, the coordinator can coordinate each model training participating device corresponding to other model training tasks to be completed to perform federated learning modeling, thereby avoiding the need for the coordinator to perform computing tasks and consume computing resources when each federated participant performs local iterative training. In this way, the purpose of making full use of the coordinator's computing resources is achieved, and the utilization rate of the coordinator's computing resources is improved. Therefore, the technical problem of the low utilization rate of the coordinator's computing resources in the federated learning system is solved.

Further, referring to FIG. 2 , based on the first embodiment of the present application, in another embodiment of the present application, the federated learning modeling method is applied to a second device, and the federated learning modeling method includes:

Step C10, interact with the first device, determine model training information, and obtain device status information, so as to determine whether to participate in the to-be-completed model training task corresponding to the model training information based on the device status information;

In this embodiment, the model training information includes model index information and model training time information, and the device status information includes available computing resources of the second device, where the available computing resources are the second device The computing resources that can be invoked during the model training time period corresponding to the model training task to be completed.

Before performing step C10, the second device negotiates and interacts with the first device to determine each model training task to be completed.

Interact with the first device, determine model training information, and obtain device status information, so as to determine whether to participate in the to-be-completed model training task corresponding to the model training information based on the device status information. The first device performs negotiation and interaction, obtains model training information, and determines available computing resources, and then judges whether the available computing resources satisfy the model training tasks to be completed corresponding to the model training information, and if the available computing resources satisfy the to-be-completed model training tasks Complete the model training task, then determine to participate in the to-be-completed model training task, and if the available computing resources do not meet the to-be-completed model training task, determine not to participate in the to-be-completed model training task, for example, assuming the to-be-completed model training task Completing the model training task requires 50% of all computing resources of the second device, and the available computing resources that can be invoked by the second device are 40%, then the available computing resources do not satisfy the model training task to be completed. , and then determine not to participate in the to-be-completed model training task.

Step C20, if participating in the to-be-completed model training task, execute a preset federated learning modeling process by coordinating and interacting with the first device to complete the to-be-completed model training task.

In this embodiment, if participating in the to-be-completed model training task, the preset federated learning modeling process is executed by coordinating and interacting with the first device to complete the to-be-completed model training task. Specifically, If participating in the to-be-completed model training task, determine the to-be-trained model corresponding to the to-be-completed model training task, and perform iterative training on the to-be-trained model until the to-be-trained model reaches the preset number of iterative training times, then Acquire the local model parameters of the model to be trained after iterative training, and send the local model parameters to the first device for the first device to be based on the local model parameters sent by each of the second devices, Calculate the latest federated model parameters, and broadcast the latest federated model parameters to each of the second devices, and then the second device receives the latest federated model parameters, and updates the to-be-to-be-modeled parameters based on the latest federated model parameters Train the model, and judge whether the updated model to be trained satisfies the preset iteration end condition. If the updated model to be trained satisfies the preset iteration end condition, it is judged that the model training task to be completed is completed. If the updated model to be trained satisfies the preset iteration end condition If the model does not meet the preset iteration end condition, the model to be trained is re-iteratively trained for the first device to recalculate the latest federated model parameters to re-update the model to be trained until the update The final model to be trained satisfies the preset iteration end condition.

Wherein, in step C20, the step of executing the preset federated learning modeling process by coordinating and interacting with the first device includes:

Step C21: Determine the to-be-trained model corresponding to the to-be-completed model training task, and perform iterative training on the to-be-trained model until the to-be-trained model reaches a preset number of iterations, and obtain a local model corresponding to the to-be-trained model parameter;

In this embodiment, the to-be-trained model corresponding to the to-be-completed model training task is determined, and the to-be-trained model is iteratively trained until the to-be-trained model reaches a preset number of iterations, and the corresponding to-be-trained model is obtained. The local model parameters of The updated local model parameters of the model to be trained.

Step C22, sending the local model parameters to the first device for the first device to calculate the latest federated model parameters based on the local model parameters;

In this embodiment, the local model parameters are sent to the first device, so that the first device can calculate the latest federated model parameters based on the local model parameters, and specifically, send the local model parameters to the first device, so that the first device can calculate the latest federated model parameters corresponding to each of the local model parameters based on the local model parameters sent by the associated second devices through preset aggregation rules, wherein the The preset aggregation rules include weighted averaging, summation, and the like.

Step C23: Receive the latest federated model parameters fed back by the first device, and based on the latest federated model parameters, update the to-be-trained model until the local model reaches a preset training end condition, and obtain the to-be-completed model The target modeling model corresponding to the training task.

In this embodiment, the latest federated model parameters fed back by the first device are received, and based on the latest federated model parameters, the model to be trained is updated until the local model reaches a preset training end condition, and the The target modeling model corresponding to the model training task to be completed, specifically, receiving the latest federated model parameters fed back by the first device, and replacing and updating the local model parameters in the to-be-trained model with the latest federated model parameters, Obtain the replaced and updated model to be trained, and determine whether the replaced and updated model to be trained satisfies the preset iterative training end condition, and if the replaced and updated model to be trained satisfies the preset iterative training end condition, then replace The updated model to be trained is used as the target modeling model. If the replaced and updated model to be trained does not meet the preset iterative training end condition, then the iterative training is performed on the model to be trained again, so that the The to-be-trained model is replaced and updated until the replaced and updated model to-be-trained meets the preset iterative training end condition.

In this embodiment, by interacting with the first device, the model training information is determined, and the device status information is obtained, so as to determine whether to participate in the to-be-completed model training task corresponding to the model training information based on the device status information, and further if Participate in the to-be-completed model training task, execute a preset federated learning modeling process by coordinating and interacting with the first device to complete the to-be-completed model training task. That is, this embodiment provides a modeling method based on federated learning, that is, before performing federated learning modeling, it is determined whether to participate in any The model training task to be completed corresponding to the model training information, and if it is determined to participate, it can coordinate and interact with the first device, and execute the preset federated learning modeling process to complete the model training task, that is, all the The second device can independently choose to participate in the to-be-completed model training task before each federated learning modeling, thereby laying a foundation for solving the technical problem of low utilization of the coordinator's computing resources in the federated learning system.

Referring to FIG. 3 , FIG. 3 is a schematic diagram of the device structure of the hardware operating environment involved in the solution of the embodiment of the present application.

As shown in FIG. 3 , the federated learning modeling device may include: a processor 1001 , such as a CPU, a memory 1005 , and a communication bus 1002 . Among them, the communication bus 1002 is used to realize the connection communication between the processor 1001 and the memory 1005 . The memory 1005 may be high-speed RAM memory, or may be non-volatile memory, such as disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001 .

Optionally, the federated learning modeling device may further include a rectangular user interface, a network interface, a camera, an RF (Radio Frequency, radio frequency) circuit, a sensor, an audio circuit, a WiFi module, and the like. The rectangular user interface may include a display screen (Display), an input sub-module such as a keyboard (Keyboard), and the optional rectangular user interface may also include a standard wired interface and a wireless interface. Optional network interfaces may include standard wired interfaces and wireless interfaces (eg, WI-FI interfaces).

Those skilled in the art can understand that the structure of the federated learning modeling device shown in FIG. 3 does not constitute a limitation on the federated learning modeling device, and may include more or less components than those shown in the figure, or combine some components, Or a different component arrangement.

As shown in FIG. 3 , the memory 1005 as a computer storage medium may include an operating system, a network communication module and a federated learning modeling program. The operating system is a program that manages and controls the hardware and software resources of the federated learning modeling device, and supports the operation of the federated learning modeling program and other software and/or programs. The network communication module is used to realize the communication between various components in the memory 1005, and communicate with other hardware and software in the federated learning modeling system.

In the federated learning modeling device shown in FIG. 3 , the processor 1001 is configured to execute the federated learning modeling program stored in the memory 1005 to implement the steps of the federated learning modeling method described in any one of the above.

The specific implementation manner of the federated learning modeling device of the present application is basically the same as the above-mentioned embodiments of the federated learning modeling method, and details are not repeated here.

An embodiment of the present application further provides a federated learning modeling apparatus, where the federated learning modeling apparatus is applied to the first device, and the federated learning modeling apparatus includes:

A negotiation module, configured to negotiate and interact with each second device associated with the first device, determine each model training task to be completed, and determine in each of the second devices the corresponding model training tasks to be completed respectively Participating equipment for each model training;

A coordination module, configured to obtain the model training time periods corresponding to the model training tasks to be completed, and coordinate the model training participating devices corresponding to the model training tasks to be completed based on the model training time periods A preset federated learning modeling process is performed to complete each of the to-be-completed model training tasks.

Optionally, the negotiation module includes:

an obtaining unit, configured to obtain model training information corresponding to each of the model training tasks;

A determination unit, configured to determine each of the model training participating devices corresponding to each of the to-be-completed model training tasks, based on each of the model training information, by performing a willingness confirmation interaction with each of the second devices.

Optionally, the determining unit includes:

A first sending subunit, configured to send each of the model index information to each of the second devices, so that each of the second devices can, based on the acquired model training requirement information and each of the model index information, respectively, in the Determining each target model training task involved in each of the model training tasks, and generating first determination information corresponding to each of the target model training tasks;

The first determination subunit is configured to determine each of the model training participating devices corresponding to each of the to-be-completed model training tasks, based on each of the first determination information respectively fed back by each of the second devices.

Optionally, the determining unit further includes:

The second sending subunit is configured to send each of the model training time information to each of the second devices respectively, so that each of the second devices can use the acquired training time limit information and each of the model training time information respectively. , determine each target model training task involved in each of the model training tasks, and generate second determination information corresponding to each of the target model training tasks;

A second determination subunit, configured to determine each of the model training participating devices corresponding to each of the to-be-completed model training tasks based on each of the second determination information fed back by each of the second devices.

Optionally, the coordination module includes:

a computing unit, configured to respectively receive local model parameters sent by each of the model training participating devices corresponding to the model training time period in each of the model training time periods, and calculate the latest federated model parameters based on a preset aggregation rule ;

a first determination unit, configured to determine whether the latest federated model parameters meet the preset training task end condition;

An update unit, configured to send the latest federated model parameters to each of the second devices if the latest federated model parameters satisfy the preset training task end condition, so that each of the second devices can update their respective local model;

A second determination unit, configured to send the latest federated model parameters to each of the model training participating devices, respectively, if the latest federated model parameters do not meet the preset training task ending condition, for each of the The participating devices update their respective federated participation models to recalculate the latest federated model parameters until the newest federated model parameters satisfy the preset training task end condition.

The specific implementation of the federated learning modeling apparatus of the present application is basically the same as the above-mentioned embodiments of the federated learning modeling method, and details are not described herein again.

To achieve the above purpose, an embodiment of the present application further provides a federated learning modeling device, the federated learning modeling device is applied to the second device, and the federated learning modeling device includes:

an interaction module, configured to interact with the first device, determine model training information, and obtain device status information, so as to determine whether to participate in the to-be-completed model training task corresponding to the model training information based on the device status information;

The federated learning modeling module is configured to perform a preset federated learning modeling process by coordinating and interacting with the first device if participating in the to-be-completed model training task to complete the to-be-completed model training task.

Optionally, the federated learning modeling module includes:

an iterative training unit, configured to determine the to-be-trained model corresponding to the to-be-completed model training task, perform iterative training on the to-be-trained model until the to-be-trained model reaches a preset number of iterations, and obtain the corresponding to-be-trained model The local model parameters of ;

a sending unit, configured to send the local model parameters to the first device, so that the first device can calculate the latest federated model parameters based on the local model parameters;

an update unit, configured to receive the latest federated model parameters fed back by the first device, and based on the latest federated model parameters, update the to-be-trained model until the local model reaches a preset training end condition, and obtain the to-be-trained model Complete the target modeling model corresponding to the model training task.

The specific implementation of the federated learning modeling apparatus of the present application is basically the same as the above-mentioned embodiments of the federated learning modeling method, and details are not described herein again.

An embodiment of the present application provides a readable storage medium, and the readable storage medium stores one or more programs, and the one or more programs can also be executed by one or more processors to implement The steps of the federated learning modeling method described in any one of the above.

The specific implementations of the readable storage medium of the present application are basically the same as the above-mentioned embodiments of the federated learning modeling method, and are not repeated here.

The above are only the preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present application, or directly or indirectly applied in other related technical fields , are similarly included within the scope of patent processing of this application.

Claims (10)

1. The federated learning modeling method is applied to first equipment, and comprises the following steps:

negotiating and interacting with each second device associated with the first device, determining each model training task to be completed, and determining each model training participating device corresponding to each model training task to be completed in each second device;

and obtaining a model training time period corresponding to each to-be-completed model training task, and coordinating each model training participating device corresponding to each to-be-completed model training task to perform a preset federal learning modeling process based on each model training time period so as to complete each to-be-completed model training task.

2. The federal learning modeling method as claimed in claim 1, wherein the step of determining, in each of the second devices, each model training participant device corresponding to each of the to-be-completed model training tasks includes:

obtaining model training information corresponding to each model training task;

and determining each model training participation device corresponding to each model training task to be completed by carrying out intention confirmation interaction with each second device based on each model training information.

3. The federated learning modeling method of claim 2, wherein the model training information includes model index information,

the step of determining each model training participating device corresponding to each to-be-completed model training task by performing intention confirming interaction with each second device based on each model training information includes:

respectively sending the model index information to each second device, so that each second device determines each target model training task participating in each model training task based on the acquired model training demand information and each model index information, and generates first determination information corresponding to each target model training task;

and determining each model training participating device corresponding to each model training task to be completed based on each first determination information fed back by each second device.

4. The federal learning modeling method as claimed in claim 2, wherein the model training information includes model training time information,

the step of determining each model training participating device corresponding to each to-be-completed model training task by performing intention confirming interaction with each second device based on each model training information includes:

respectively sending the model training time information to each second device, so that each second device determines each target model training task participating in each model training task based on the acquired training time limit information and each model training time information, and generates second determination information corresponding to each target model training task;

and determining each model training participating device corresponding to each model training task to be completed based on each piece of second determination information fed back by each piece of second equipment.

5. The federal learning modeling method as claimed in claim 1, wherein the step of coordinating each model training participating device corresponding to each model training task to be completed to perform a preset federal learning modeling procedure based on each model training time period comprises:

in each model training time period, receiving local model parameters sent by each model training participating device corresponding to the model training time period respectively, and calculating the latest federal model parameters based on a preset aggregation rule;

determining whether the latest federal model parameters meet preset training task end conditions;

if the latest federal model parameters meet the preset training task end conditions, the latest federal model parameters are sent to the second devices so that the second devices can update respective local models;

and if the latest federal model parameters do not meet the preset training task end conditions, respectively sending the latest federal model parameters to each model training participatory device, so that each model participatory device updates the respective federal participatory model, and recalculating the latest federal model parameters until the latest federal model parameters meet the preset training task end conditions.

6. The federated learning modeling method is applied to second equipment, and comprises the following steps:

interacting with the first equipment, determining model training information, and acquiring equipment state information so as to determine whether to participate in a to-be-completed model training task corresponding to the model training information based on the equipment state information;

and if the model training task to be completed is participated, executing a preset federal learning modeling process by coordinating and interacting with the first equipment so as to complete the model training task to be completed.

7. The federal learning modeling method as claimed in claim 6, wherein said step of executing a predetermined federal learning modeling procedure through a coordinated interaction with said first device comprises:

determining a model to be trained corresponding to the model training task to be completed, and performing iterative training on the model to be trained until the model to be trained reaches a preset iteration number, and acquiring local model parameters corresponding to the model to be trained;

sending the local model parameters to the first device, so that the first device can calculate the latest federal model parameters based on the local model parameters;

and receiving the latest federal model parameters fed back by the first equipment, updating the model to be trained based on the latest federal model parameters until the local model reaches a preset training end condition, and obtaining a target modeling model corresponding to the model training task to be completed.

8. The utility model provides a federal study modeling device, its characterized in that, federal study modeling device includes:

the negotiation module is used for carrying out negotiation interaction with each second device associated with the first device, determining each model training task to be completed, and determining each model training participating device corresponding to each model training task to be completed in each second device;

and the coordination module is used for acquiring model training time periods corresponding to the model training tasks to be completed, coordinating each model training participating device corresponding to each model training task to be completed to carry out a preset federal learning modeling process based on each model training time period, so as to complete each model training task to be completed.

9. The federal learning modeling apparatus is characterized in that the federal learning modeling apparatus includes: a memory, a processor, and a program stored on the memory for implementing the federated learning modeling method,

the memory is used for storing a program for realizing the federal learning modeling method;

the processor is configured to execute a program implementing the federal learning modeling method to implement the steps of the federal learning modeling method as claimed in any of claims 1 to 5 or 6 to 7.

10. A readable storage medium having stored thereon a program for implementing a federal learning modeling method, the program being executed by a processor to implement the steps of the federal learning modeling method as claimed in any of claims 1 to 5 or 6 to 7.

Images