Training method and device of federated learning model, electronic equipment, medium and product

Abstract

The application provides a training method and device of a federated learning model, electronic equipment, a computer readable storage medium and a computer program product; the method is applied to an active party device with a label, and comprises the following steps: obtaining a sample identifier of a sample used for training the federated learning model and a first sample feature of a locally stored sample; sending a sample determination request carrying the sample identifier to at least one passive party device; wherein the sample determination request is used for the passive party device to filter a second sample feature corresponding to the sample identifier from a local sample feature library; receiving response information corresponding to the sample determination request sent by the passive party device; when the response information indicates that the filtering of the second sample feature corresponding to the sample identifier is completed, the federated learning model is trained in combination with the first sample feature and the second sample feature of the sample. Through the application, the accuracy of the filtered training sample can be improved, and the modeling efficiency of the federated learning system can be effectively improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a training method, apparatus, electronic device, computer-readable storage medium, and computer program product for a federated learning model. Background Technology

[0002] Federated learning is an emerging privacy-preserving technology that enables the effective collaboration of data from various parties for model training without leaving the local machine.

[0003] During the training of related federated learning models, a large amount of information needs to be manually configured, such as table names and namespaces obtained through brute-force attacks. This exposes a large number of system technical details unrelated to model training to users. Ordinary modeling users need to carefully read the documentation and receive guidance from professionals before they can carry out federated modeling. Since brute-force attacks and modeling in federated modeling are time-consuming, any errors in manual configuration will waste a lot of time, resulting in low modeling efficiency. Summary of the Invention

[0004] This application provides a training method, apparatus, electronic device, computer-readable storage medium, and computer program product for a federated learning model, which can improve the accuracy of selecting training samples and effectively improve the modeling efficiency of the federated learning system.

[0005] The technical solution of this application embodiment is implemented as follows:

[0006] This application provides a method for training a federated learning model, applied to a labeled active-party device, comprising:

[0007] Obtain the sample identifier of the sample used to train the federated learning model, and the first sample feature of the sample stored locally;

[0008] Send a sample determination request carrying the sample identifier to at least one passive party device;

[0009] The sample determination request is used by the passive device to filter and obtain a second sample feature corresponding to the sample identifier from the local sample feature library;

[0010] Receive response information from the passive device corresponding to the sample determination request;

[0011] When the response information indicates that the second sample feature filtering corresponding to the sample identifier is completed, the federated learning model is trained by combining the first sample feature and the second sample feature of the sample.

[0012] This application provides a training apparatus for a federated learning model, comprising:

[0013] The acquisition module is used to acquire the sample identifier of the sample used to train the federated learning model, and the first sample feature of the sample stored locally.

[0014] A sending module is configured to send a sample determination request carrying the sample identifier to at least one passive party device; wherein, the sample determination request is used by the passive party device to filter and obtain a second sample feature corresponding to the sample identifier from a local sample feature library;

[0015] A receiving module is used to receive response information sent by the passive device in response to the sample determination request;

[0016] The training module is used to train the federated learning model by combining the first sample feature and the second sample feature of the sample when the response information indicates that the second sample feature filtering corresponding to the sample identifier is completed.

[0017] In the above scheme, the acquisition module is further used to acquire sample configuration information, which includes sample feature range and sample identifier. The sample feature range is used to indicate the range to which the sample features of the samples used to train the federated learning model belong.

[0018] Based on the sample identifier, determine the local sample features corresponding to the sample identifier in the local sample feature library;

[0019] Based on the range of sample features, sample features that fall within the range of sample features are selected as the first sample features.

[0020] In the above scheme, the acquisition module is further used to acquire sample configuration information, the sample configuration information including sample feature range;

[0021] Based on the range of sample features, select sample features from the local sample feature library that fall within the range of sample features and use them as the first sample features.

[0022] The sample identifier of the first sample feature is used as the sample identifier of the sample used to train the federated learning model.

[0023] In the above scheme, the sample configuration information includes at least one passive party device identifier, and the sending module is further configured to send a sample determination request carrying the sample identifier to the corresponding passive party device according to the at least one passive party device identifier.

[0024] In the above scheme, the sending module is further configured to determine multiple candidate participant devices that can collaboratively train the federated learning model with the active party device;

[0025] Determine the priority of each of the candidate participant devices;

[0026] Based on the priority, at least one candidate participant device is selected from the plurality of passive party devices as the passive party device.

[0027] In the above scheme, the training module is further configured to, when the first sample feature and the second sample feature belong to at least two feature ranges, combine the first sample feature and the second sample feature in each feature range respectively to train the federated learning model, so as to obtain a federated learning model corresponding to each feature range;

[0028] The prediction accuracy of the federated learning model corresponding to each of the aforementioned feature ranges is evaluated to obtain the evaluation results;

[0029] The evaluation results are output using the target output method.

[0030] In the above scheme, the sending module is further configured to obtain the sending order of the at least two passive devices when the number of passive devices is at least two.

[0031] According to the sending order, a sample determination request carrying the sample identifier is sent to the passive device one by one;

[0032] During the process of sending the sample determination request to each of the passive party devices, when it is determined that the second sample feature corresponding to the sample identifier has been successfully filtered based on the response information returned by each of the passive party devices, the sending of the sample determination request is stopped.

[0033] In the above scheme, the training module is further configured to, when the number of sample identifiers is at least two, and the response information indicates that the second sample feature filtering corresponding to the first sample identifier among the at least two sample identifiers has failed,

[0034] Output the sample determination failure information corresponding to the first sample identifier to indicate that the second sample feature filtering corresponding to the first sample identifier has failed.

[0035] In the above scheme, the training module is further configured to determine the successfully filtered sample identifier as the second sample identifier when the response information indicates that the second sample feature filtering corresponding to the sample identifier is completed and partially successful.

[0036] From a plurality of first sample features corresponding to the sample identifier, filter the first sample features corresponding to the second sample identifier;

[0037] The federated learning model is trained by combining the first sample feature corresponding to the second sample identifier and the second sample feature corresponding to the second sample identifier.

[0038] This application provides an electronic device, including:

[0039] Memory, used to store executable instructions;

[0040] The processor, when executing executable instructions stored in the memory, implements the training method of the federated learning model provided in the embodiments of this application.

[0041] This application provides a computer-readable storage medium storing executable instructions for inducing a processor to execute and implement the training method of the federated learning model provided in this application.

[0042] This application provides a computer program product, including a computer program that, when executed by a processor, implements the training method of the federated learning model provided in this application.

[0043] The embodiments of this application have the following beneficial effects:

[0044] Compared to training federated learning models by directly using all sample features from the local feature library, this embodiment sends a sample determination request carrying local sample identifiers to the passive device and receives the corresponding response from the passive device. This determines the sample feature data to participate in the federated model training, and the federated learning model is trained based on the determined sample feature data. This improves the accuracy of selecting training samples, effectively increases the modeling efficiency of the federated learning system, and enhances the user experience. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the architecture of the training system for the federated learning model provided in this application embodiment;

[0046] Figure 2 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application;

[0047] Figure 3 This is a flowchart illustrating the training method of the federated learning model provided in the embodiments of this application;

[0048] Figure 4 This is a schematic diagram of information interaction in the federated learning model provided in an embodiment of this application;

[0049] Figure 5 This is a schematic diagram of a method for obtaining sample identifiers and sample features provided in an embodiment of this application;

[0050] Figure 6A This is a schematic diagram of the visual interface for setting sample configuration information provided in the embodiments of this application;

[0051] Figure 6B This is a schematic diagram of the text content of the sample configuration information settings provided in the embodiments of this application;

[0052] Figure 6C This is another visual interface diagram of the sample configuration information settings provided in the embodiments of this application;

[0053] Figure 7 This is a schematic diagram of a method for obtaining sample features provided in an embodiment of this application;

[0054] Figure 8 This is a schematic diagram of a method for determining a passive device provided in an embodiment of this application;

[0055] Figure 9 This is a schematic diagram of a method for sending a sample determination request provided in an embodiment of this application;

[0056] Figures 10A-10C This is a schematic diagram of the response information provided in this application.

[0057] Figure 11 This is a schematic diagram of the training method of the federated learning model provided in the embodiments of this application;

[0058] Figure 12 This is a schematic diagram of a method for obtaining sample features provided in an embodiment of this application;

[0059] Figure 13 This is a schematic diagram of a risk control model based on vertical federated learning.

[0060] Figure 14 This is a training method for a risk control model based on federated learning provided in the embodiments of this application. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0062] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0063] If the application documents contain similar descriptions such as "first / second", the following explanation shall be added: In the following description, the terms "first / second / third" are used only to distinguish similar objects and do not represent a specific order of objects. It is understood that "first / second / third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0064] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0065] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0066] 1) Federated machine learning (also known as federated learning, joint learning, or consortium learning) is a machine learning framework that effectively helps multiple organizations use data and perform machine learning modeling while meeting user privacy, data security, and government regulations. As a distributed machine learning paradigm, federated learning effectively solves the data silo problem, allowing participants to jointly model data without sharing it, thus technically breaking down data silos and achieving collaborative intelligence.

[0067] Federated learning (aka Federated Machine Learning) can be divided into three categories: Horizontal Federated Learning, Vertical Federated Learning, and Federated Transfer Learning.

[0068] Horizontal federated learning, also known as feature-aligned federated learning, involves participants whose data features are aligned. It is suitable for situations where participants' data features overlap significantly, while their sample identifiers (IDs, identity documents) overlap less. Vertical federated learning, also known as sample-aligned federated learning, involves participants whose training samples are aligned. It is suitable for situations where participants' training sample IDs overlap significantly, while their data features overlap less.

[0069] 2) Credential stuffing: Based on a large amount of user data, this method uses users' similar registration habits (same username and password) to attempt to log in to other websites. Simply put, it's like copying a bunch of keys from someone else and then trying to open different doors in a building.

[0070] In the vertical federated learning model, it can be understood that the active party device sends a brute-force request (or sample determination request) to other passive party devices based on the sample identifiers in its local sample library. The passive party devices then filter the sample features corresponding to the sample identifiers in their local databases.

[0071] 3) Risk control: This refers to risk control achieved through risk assessment of business operations, such as risk control of bank loan funds, including pre-loan risk assessment (such as credit assessment), risk monitoring during the loan process (such as monitoring credit changes), and post-loan loan management (such as refinancing based on credit changes).

[0072] In the modeling process of risk control, recommendation and other problems, the training techniques of related federated learning models expose a large number of system technical details that are irrelevant to modeling to users, resulting in low modeling efficiency and poor user experience.

[0073] Based on this, embodiments of this application provide a training method, apparatus, electronic device, computer-readable storage medium, and computer program product for a federated learning model, which can improve the accuracy of selecting training samples and effectively improve the modeling efficiency of the federated learning system.

[0074] First, the training system for the federated learning model provided in this application embodiment will be described, see [link to relevant documentation]. Figure 1 , Figure 1This is a schematic diagram of the architecture of a training system for a federated learning model provided in this application embodiment. In the training system 100 of the federated learning model, the passive party device 400 (passive party device 400-1 and passive party device 400-2 are shown as examples) is connected to the active party device 200 through a network 300. The network 300 can be a wide area network, a local area network, or a combination of both, and data transmission is achieved using a wireless link. In some embodiments, the passive party device 400 can be a laptop, tablet, desktop computer, smartphone, dedicated messaging device, portable gaming device, smart speaker, smartwatch, etc., but is not limited to these. The active party device 200 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The passive device 400 and the active device 200 can be connected directly or indirectly through wired or wireless communication, and this application embodiment does not impose any restrictions.

[0075] The active device 200 is used to acquire sample identifiers of samples used for training the federated learning model, and first sample features of samples stored locally; send a sample determination request carrying the sample identifier to at least one passive device; wherein, the sample determination request is used by the passive device to filter and obtain second sample features corresponding to the sample identifier from the local sample feature library; receive response information of the corresponding sample determination request sent by the passive device; when the response information indicates that the second sample feature of the corresponding sample identifier has been filtered, the federated learning model is trained by combining the first sample feature and the second sample feature of the sample.

[0076] The passive device 400 is used to receive a sample determination request carrying a sample identifier sent by the active device, parse the sample determination request to obtain the sample identifier; based on the sample identifier, it filters the second sample feature corresponding to the sample identifier from the local sample feature library, and sends the response information corresponding to the sample determination request to the active device.

[0077] See Figure 2 , Figure 2 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. In practical applications, the electronic device 500 can be implemented as follows: Figure 1 The passive device 400 or the active device 200 in the middle are electronic devices. Figure 1 Taking the active device 200 shown as an example, an electronic device implementing the training method of the federated learning model of the present application embodiment will be described. Figure 2The illustrated electronic device 500 includes at least one processor 510, a memory 550, at least one network interface 520, and a user interface 530. The various components in the electronic device 500 are coupled together via a bus system 540. It is understood that the bus system 540 is used to implement communication between these components. In addition to a data bus, the bus system 540 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 2 The general labeled all buses as Bus System 540.

[0078] The processor 510 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0079] User interface 530 includes one or more output devices 531 that enable the presentation of media content, including one or more speakers and / or one or more visual displays. User interface 530 also includes one or more input devices 532, including user interface components that facilitate user input, such as a keyboard, mouse, microphone, touch screen display, camera, other input buttons and controls.

[0080] The memory 550 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 550 may optionally include one or more storage devices physically located away from the processor 510.

[0081] The memory 550 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 550 described in this application embodiment is intended to include any suitable type of memory.

[0082] In some embodiments, memory 550 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0083] Operating system 551 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks;

[0084] The network communication module 552 is used to reach other computing devices via one or more (wired or wireless) network interfaces 520, exemplary network interfaces 520 including: Bluetooth, WiFi, and Universal Serial Bus (USB), etc.

[0085] Presentation module 553 is configured to enable the presentation of information (e.g., a user interface for operating peripheral devices and displaying content and information) via one or more output devices 531 (e.g., a display screen, a speaker, etc.) associated with user interface 530;

[0086] The input processing module 554 is used to detect and translate one or more user inputs or interactions from one or more input devices 532.

[0087] In some embodiments, the training apparatus for the federated learning model provided in this application can be implemented in software. Figure 2 A training apparatus 555 for a federated learning model stored in memory 550 is shown. This apparatus can be software in the form of programs and plug-ins, and includes the following software modules: an acquisition module 5551, a transmission module 5552, a reception module 5553, and a training module 5554. These modules are logically connected and can therefore be arbitrarily combined or further divided according to their implemented functions. The functions of each module will be described below.

[0088] In other embodiments, the training device for the federated learning model provided in this application can be implemented in hardware. As an example, the training device for the federated learning model provided in this application can be a processor in the form of a hardware decoding processor, which is programmed to execute the training method of the federated learning model provided in this application. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0089] The training method of the federated learning model provided in this application will be described by referring to exemplary applications and implementations of the active-side device provided in the embodiments of this application. See also Figure 3 , Figure 3 This is a flowchart illustrating the training method of the federated learning model provided in this application embodiment, which will be combined with... Figure 3 The steps shown are explained.

[0090] In step 101, the active device acquires the sample identifiers of the samples used to train the federated learning model, as well as the first sample features of the samples stored locally.

[0091] In some embodiments, see Figure 4 , Figure 4 This diagram illustrates the information interaction of a federated learning model according to an embodiment of this application. The participating devices in the federated learning model shown in the diagram may include one active device and n (n≥1 and n is an integer) passive devices. The active device acts as the label holder and feature provider for the samples during training the federated learning model, while the passive devices only act as feature providers for the samples. The active device sends a sample determination request carrying a sample identifier to the n passive devices, enabling the passive devices to provide sample features corresponding to the sample identifier based on the sample determination request. After responding to the sample determination request, the passive devices return response information to the active device, informing the active device of their response status (successful or failed).

[0092] For example, when a federated learning model is applied to risk management or risk prediction, the labels of the training samples used to train the model can be classification labels identifying user creditworthiness, user loyalty, or user quality. Here, the active party can be a platform that conducts credit assessments, such as a financial lending platform (e.g., a bank), and the passive party can be a platform holding user feature data, such as various types of consumer platforms (e.g., shopping malls, supermarkets, online shopping platforms, gaming platforms, mobile app stores, etc.) or internet companies. The multiple data features used to train the federated learning model can include, but are not limited to, user basic information, business data, behavioral data, and device information. Basic information includes, but is not limited to, the user's name, gender, age, education, occupation, mobile phone number, ID number, address, and income data (e.g., annual salary, monthly salary). Business data includes data from multiple business scenarios. For example, if the business scenario is loan business, the business data includes loan data and repayment data. Loan data includes loan amount, loan time and lending institution (e.g., xx bank), etc. Repayment data includes repayment amount, repayment time, repayment institution and overdue repayment data (e.g., number of overdue payments and number of overdue days), etc.

[0093] In some embodiments, the active-side device can determine the sample identifier of the samples used to train the federated learning model, and the sample features provided by its own local sample feature library, based on preset sample configuration information (the sample features provided by the active-side device for training the federated learning model can be referred to as the first sample features). That is, the sample features provided by the active-side device are called the first sample features. (See reference...) Figure 5 , Figure 5 This is a schematic diagram illustrating the method for obtaining sample identifiers and sample features provided in an embodiment of this application. Based on Figure 3 , Figure 3 Step 101 shown is implemented through steps 1011a to 1013a, and will be explained in conjunction with each step.

[0094] Step 1011a: The active device acquires sample configuration information, which includes the sample feature range and sample identifier.

[0095] In practice, the active device can obtain sample configuration information set by the user by providing a visual information configuration interface. (See also...) Figure 6A , Figure 6A This is a schematic diagram of the visual interface for setting sample configuration information provided in this application embodiment. The sample configuration information that can be set by the user may include the active party device identifier (number 1), the passive party device identifier (number 2), the coordinating party device identifier (number 3), the sample identifier (number 4), and the sample feature range (number 5). The active party device identifier refers to the device identifier of the participating party device carrying tag information. The active party device is the active party (initiator) for training and using the federated learning model. In practical applications, there is usually only one active party in the federated learning model, and the active party device identifier can be the default. The passive party device identifier is the device identifier of the feature provider device that provides sample features to the federated learning model. The passive party device establishes communication with the active party device through the network and collaborates with the active party device to train the federated learning model. In practical applications, there can be multiple passive party devices, and the user can set the range of samples. Figure 6A The input boxes corresponding to the passive party device identifiers shown in the diagram can be used to enter multiple passive party device identifiers. The coordinator device identifier is the device identifier used by the federated learning model to aggregate parameters and send the aggregated parameters to each participating party device. In practical applications, the active party can also function as the coordinator; that is, the active party device and the coordinator device can be the same server device. It should be noted that in practical applications… Figure 6AThe active device identifier, passive device identifier, and coordinator device identifier shown can all be from the same server device. The sample identifier is determined by the active device as a sample to participate in the federated learning model. There can be multiple sample identifiers. The active device can determine the local sample features corresponding to the sample identifier in its local sample feature library. It can also send a sample determination request carrying the sample identifier to the passive device, requesting the passive device to provide the sample features corresponding to the sample identifier. The sample feature range refers to the range to which the sample features belong. Based on the sample feature range, the active device can determine the sample features that fall within the range, which are then used as the sample features corresponding to the active device actually participating in the federated learning model. In practical applications, the sample feature range can be a feature range described by time information, such as the sample feature range value information shown in Figure 5, which includes {last month, last three months, last six months, last year, custom range}, etc. The value information for the sample feature range can be set according to actual conditions; this application embodiment does not restrict the value of the sample feature range.

[0096] For example, taking a risk control model based on vertical federated learning as an example, user U initiates a loan application to the initiating party (bank A). The initiating party needs to obtain user U's credit information (which can be determined by the number of overdue payments) from other participating parties such as banks, insurance companies, and supermarkets, and then determine whether to grant a loan to user U and the loan amount. The time T when user U initiates the loan application can be used as a time point to obtain the feature information of user U over the most recent month, starting from time T (mainly the number of overdue payments provided by the passive party).

[0097] In practice, the active device can configure sample-related information for the federated learning model through a sample configuration information file. (See [link to relevant documentation]). Figure 6B , Figure 6B This is a schematic diagram of the text content of the sample configuration information settings provided in the embodiments of this application, including... Figure 6A The sample configuration information shown also provides some information about federated learning models, which can be directly used by the user. Figure 6A The visualized interface for sample configuration information provides more efficient sample configuration and a better user experience.

[0098] Step 1012a: Based on the sample identifier, determine the local sample feature library and the local sample feature corresponding to the sample identifier.

[0099] In practice, the active device selects local sample features corresponding to the sample identifiers from the local sample feature library based on the identified sample identifiers that need to participate in the training of the federated learning model.

[0100] Step 1013a: Based on the sample feature range, select the sample features that fall within the sample feature range and use them as the first sample features.

[0101] In actual implementation, after the active device obtains the local sample features corresponding to the sample identifier through step 1012a, it further filters the local sample features obtained in step 1012a according to the set sample feature range. At this time, the sample features corresponding to the sample identifier and falling within the sample feature range in the local feature library corresponding to the active device are obtained. The sample features corresponding to the sample identifier after the second filtering are used as the first sample features of the federated learning model.

[0102] In some embodiments, the sample configuration information sets sample identifiers, which by default uses all sample identifiers from the active device's local sample feature library whose sample features fall within the sample feature range. (See reference) Figure 7 , Figure 7 This is a schematic diagram of a method for obtaining sample features provided in an embodiment of this application. Based on Figure 3 , Figure 3 Step 101 shown is implemented through steps 1011b to 1013b, which will be explained in conjunction with each step.

[0103] Step 1011b: The active device acquires sample configuration information, which includes the sample feature range.

[0104] In actual implementation, see Figure 6C , Figure 6C This is another visual interface diagram illustrating the sample configuration information settings provided in this application embodiment. The sample configuration information may include the active party device identifier shown in number 1, the passive party device identifier shown in number 2, the coordinating party device identifier shown in number 3, the sample feature range shown in number 4c, etc., and is related to... Figure 6A In contrast, the sample configuration information settings interface does not require setting a sample identifier.

[0105] Step 1012b: Based on the sample feature range, filter the sample features in the local sample feature library that fall within the sample feature range and use them as the first sample features.

[0106] In actual implementation, the active device according to Figure 6C The system sets a range of sample features and filters sample features from the local sample feature library that fall within this range. These filtered sample features are then used as sample features provided by the active device in the training process of the federated learning model.

[0107] Step 1013b: Use the sample identifier of the first sample feature as the sample identifier of the sample used to train the federated learning model.

[0108] In practice, the active device will use the sample identifier of the sample features obtained through step 1012b as the sample identifier of the sample used to train the federated learning model.

[0109] In step 102, a sample determination request carrying a sample identifier is sent to at least one passive party device; wherein, the sample determination request is used by the passive party device to filter and obtain a second sample feature corresponding to the sample identifier from a local sample feature library.

[0110] In some embodiments, when the active device sends a sample determination request to the passive device, the passive device can be directly determined based on the passive device identifier specified in the sample configuration information. Specifically, when the sample configuration information includes at least one passive device identifier, the active device sends a sample determination request carrying a sample identifier to the corresponding passive device based on at least one passive device identifier.

[0111] In some embodiments, when multiple passive party devices exist, each can collaboratively train the same federated learning model with the active party device. However, due to the actual computing power of the active party devices and the resources allocated to training the federated learning model, the active party device may not be able to establish communication with all the passive party devices. Therefore, in practice, candidate participating devices are typically determined according to the priority of the passive party devices. See also Figure 8 , Figure 8 This is a schematic diagram of a method for determining a passive device provided in an embodiment of this application, combined with... Figure 8 The steps shown are explained.

[0112] Step 201: The active party device identifies multiple candidate participating devices that can collaborate with the active party device to train the federated learning model.

[0113] In practice, there may be multiple passive devices communicating with the active device via the network. However, each passive device and the active device may be training models suitable for different scenarios. Therefore, the active device pre-sets multiple candidate participating devices that can collaboratively train the federated learning model with itself.

[0114] Step 202: Determine the priority of each candidate participant's device.

[0115] In practice, there are several ways to determine the priority of candidate participant devices. One method is to set the priority based on the amount of sample feature data contained in the candidate participant device's local sample library; devices with more sample feature data have higher priority, and those with less sample feature data have lower priority. Alternatively, the priority can be set based on the distance between the actual geographical location of the candidate participant device and the geographical location of the initiating device; the closer the distance, the higher the priority of the corresponding device.

[0116] Step 203: Based on priority, select at least one candidate participant device from multiple passive party devices as a passive party device.

[0117] In some embodiments, when there are multiple passive devices, a sample determination request can be sent to each passive device sequentially according to a pre-set sending order for each passive device. See also Figure 9 , Figure 9 This is a schematic diagram of a method for sending a sample determination request provided in an embodiment of this application, based on... Figure 3 When there are multiple passive devices, Figure 3 Step 102 shown is implemented through steps 1021 to 1023, which will be explained in conjunction with each step.

[0118] Step 1021: The active device obtains the sending order of at least two passive devices.

[0119] In practice, there are several ways to set the sending order for the passive device. For example, the sending order can be randomly assigned. Alternatively, the sending order can be set according to the priority of the passive device; the higher the priority, the earlier it is sent.

[0120] For example, taking the determination of successful transmission through the priority of passive party devices as an example, assume that there are 6 passive party devices that can cooperate with the active party devices to train federated learning models. The passive party devices are numbered {h1, h2, ..., h6}, and the priority of the passive party devices is determined according to the amount of sample features contained in each passive party device. Let the corresponding passive party device numbers after sorting from high to low priority be {h3, h6, h2, h4, h5, h1}.

[0121] Step 1022: Send sample confirmation requests carrying sample identifiers to the passive device one by one according to the sending order.

[0122] In actual implementation, sample confirmation requests are sent to each passive device one by one according to the sending order corresponding to each passive device.

[0123] Continuing from the previous example, the passive device numbers obtained after sorting by priority from high to low are {h3, h6, h2, h4, h5, h1}. The active device sends sample confirmation requests to these 6 passive devices one by one in the aforementioned order.

[0124] Step 1023: During the process of sending the sample determination request to the passive party device, when the second sample feature corresponding to the sample identifier is successfully filtered based on the response information returned by each passive party device, the sending of the sample determination request is stopped.

[0125] In practice, the active device sends sample determination requests to each passive device in the order they are sent and receives response information from each passive device. The active device checks whether the second sample features corresponding to the sample identifiers carried in the sample determination requests have been successfully filtered. If the filtering is successful, it stops sending sample determination requests to the passive devices. It should be noted that the response information does not contain the specific attribute values ​​of the second sample features.

[0126] In step 103, the response information of the corresponding sample determination request sent by the passive device is received.

[0127] In some embodiments, the active device sends a sample determination request carrying a sample identifier to the passive device, so that the passive device returns response information in response to the sample determination request. However, given the characteristics of federated learning models, to ensure the privacy of the local data of each participant, the response information does not include feature data provided by the passive device.

[0128] The format of the received response information is described. In some embodiments, the response information may be returned in JSON format. Figures 10A-10C This is a schematic diagram of the response information provided in this application. The response information (denoted as response) in the diagram may include the passive device identifier (host_id shown in number 1), the response result code (code shown in number 2), and the sample identifier of the successfully filtered sample (result shown in number 3). Among them, host_id represents the corresponding passive device identifier; code represents the response result code; and result represents the sample identifier of the corresponding second feature data existing in the passive device.

[0129] In practice, `code` can be {-1, 0}, where `code = 0` indicates a successful response and `code = -1` indicates a failed response. To protect data privacy, `result` can be an encrypted sample identifier or a serialized string. It should be noted that when `code` returned by `host_id` is -1, `result` is null. When the number of passive devices is 1, a response containing a set of {host_id, code, result} is returned. When the number of passive devices is N (N > 1 and N is an integer), a response containing N sets of {host_id, code, result} is returned.

[0130] For example, taking a passive device with a quantity of 1 as an example, see [link to relevant documentation]. Figure 10B When the response is received, it contains a set of information including {host_id, code, result}. For example, if the number of passive devices is 3, the passive device identifiers are h3, h2, and h4. See [link / reference]. Figure 10C When the response is completed, three sets of response information containing {host_id, code, result} are returned. Among them, h3 and h2 are completed and successful, returning the sample identifier information corresponding to the sample identifier carried in the sample determination request; h4 fails and returns null.

[0131] In step 104, when the response information indicates that the second sample feature of the corresponding sample identifier has been selected, the federated learning model is trained by combining the first sample feature and the second sample feature of the sample.

[0132] In practical implementation, by combining the response information returned in step 103, the passive device that responded successfully can be identified, as well as the sample identifier that matches the sample identifier carried in the sample determination request. Therefore, the passive device can provide a second sample feature corresponding to the sample identifier, and the active device can train the federated learning model based on the first sample feature corresponding to the sample identifier provided locally and the second sample feature corresponding to the sample identifier provided by the passive device. It should be noted that the federated learning model here can be any common machine learning model suitable for federated learning scenarios, such as linear regression or linear classification. This application embodiment does not limit the specific form of the federated learning model.

[0133] In some embodiments, see Figure 11 , Figure 11 This is a schematic diagram of the training method of the federated learning model provided in the embodiments of this application, combined with Figure 11 As shown in the steps, when the first sample feature and the second sample feature belong to at least two feature ranges, step 104 can be achieved through step 104a:

[0134] Step 104a: The active device combines the first sample features and the second sample features within each feature range to train the federated learning model, thereby obtaining the federated learning model corresponding to each feature range.

[0135] In practical implementation, please refer to Figure 6A The active-side device provides a sample configuration information setting interface for users, where users only need to configure simple and easy-to-understand sample parameters. Specifically, the active-side device can segment the sample features participating in the federated learning model based on the sample feature range (segmentation according to the sample feature range). Different federated learning models with the same model structure and definition can be obtained from each segmented sample. It should be noted that by combining the first and second sample features within each feature range, the federated learning model is trained to obtain a federated learning model corresponding to each feature range; the only difference is the different sample feature ranges, resulting in federated learning models with different sample feature data.

[0136] For example, for a federated learning model M, the active device divides the sample identifier corresponding to the first sample feature into 5 segments according to the set sample feature range. Similarly, the passive device also divides the second sample feature into 5 segments. The federated learning model M is trained by combining the 5 sets of first sample features and second sample features to obtain the federated learning model M corresponding to the 5 sample feature ranges, which can be denoted as {M1, M2, M3, M4, M5}.

[0137] After step 104a is completed, the active device evaluates the accuracy of the model predictions of the federated learning model corresponding to each feature range, and outputs the evaluation results through a preset output method. The specific implementation process can be achieved through steps 105 to 106 in Figure 10:

[0138] Step 105: The active device evaluates the prediction accuracy of the federated learning model corresponding to each feature range and obtains the evaluation results.

[0139] In practice, for the same federated learning model, the active device, based on the training samples (also known as segmented samples) corresponding to each feature range, collaborates with the passive device to train the federated learning model, resulting in different federated learning models for each feature range. It's important to note that these different federated learning models are derived solely from the different feature data of the input training samples; their model definitions and structures are identical. The stability and accuracy of models trained on different segmented samples vary. The prediction values ​​obtained by inputting the training samples into the federated learning model can be compared with the model's own label information to evaluate its prediction accuracy. The closer the prediction value is to the label information, the higher the prediction accuracy of the federated learning model.

[0140] Step 106: Output the evaluation results using the target output method.

[0141] In practice, there are several ways to output the evaluation results. In some embodiments, the evaluation results can be output using a preset report template, as follows: the active device obtains the evaluation report template; the evaluation report is generated by combining the evaluation report template and the evaluation results. Alternatively, the evaluation results can be directly saved as XML or JSON files and a download interface can be provided to external users.

[0142] In some embodiments, reference Figure 12 , Figure 12 This is a schematic diagram of a method for obtaining sample features provided in an embodiment of this application. Based on Figure 3 , Figure 3 Step 104 shown is implemented through steps 1041 to 1042, and will be explained in conjunction with each step.

[0143] Step 1041: When the response information indicates that the second sample feature screening of the corresponding sample identifier is completed and partially successful, the active device determines the successfully screened sample identifier as the second sample identifier.

[0144] For example, see Figure 10C When the second sample feature filtering corresponding to each sample identifier is completed, only h3 and h2 respond successfully, returning the sample identifier information corresponding to the sample identifier carried in the sample determination request; h4 fails to respond and returns null. At this time, the active device determines the sample identifier in the result corresponding to the passive device h3 and the sample identifier in the result corresponding to the passive device h2 as the second sample identifier, that is, the second sample identifier is {uid_1, uid_2, ..., uid_n, uid_a, uid_b, ..., uid_x}.

[0145] Step 1042: Select the first sample features corresponding to the second sample identifier from multiple first sample features corresponding to the sample identifier.

[0146] Following the previous example, see [link / reference]. Figure 10C This yields multiple corresponding second sample identifiers {uid_1, uid_2, ..., uid_n, uid_a, uid_b, ..., uid_x}. The active device then selects the first sample features corresponding to each second sample identifier from its local feature library (i.e., the sample features provided by the active device during the training of the federated learning model).

[0147] Step 1043: Combine the first sample features corresponding to the second sample identifier and the second sample features corresponding to the second sample identifier to train the federated learning model.

[0148] In practice, the active device provides the first sample features corresponding to each second sample identifier from its local sample feature library, and combines them with the second sample features corresponding to each second sample identifier provided by each passive device to collaboratively train the federated learning model.

[0149] In some embodiments, a message indicating a failed sample determination request is typically broadcast to each participant in the federated learning system in the following manner. Specifically, when the number of sample identifiers is at least two, and the response information indicates that the second sample feature filtering corresponding to the first sample identifier among the at least two sample identifiers has failed, the initiating device outputs a sample determination failure message corresponding to the first sample identifier to indicate that the second sample feature filtering corresponding to the first sample identifier has failed.

[0150] In practice, when there are at least two sample identifiers, there may be a situation where at least one sample identifier fails to obtain a corresponding matching sample identifier from the response information. That is, the passive party device participating in the federated model training does not contain the second feature data corresponding to that sample identifier. In this case, the active party device will broadcast a message to the federated learning system to which the federated learning model belongs, indicating to all participants that the selection of the second sample feature corresponding to that sample identifier has failed.

[0151] In this embodiment, the active device improves the convenience of setting sample configuration information by providing a configuration interface. Based on this configuration, it performs secondary filtering of local sample features to obtain sample identifiers and corresponding first sample features for participating in the federated training model. This ensures the accuracy of the sample features provided by the active device. The active device then sends a sample confirmation request carrying the sample identifier to the passive device. Based on the response information from the passive device to this request, and combining the first sample feature corresponding to the sample identifier provided by the active device and the second sample feature corresponding to the sample identifier provided by the passive device, the federated learning model is trained. This effectively improves the modeling efficiency of the federated learning system and enhances the user experience.

[0152] The following will describe an exemplary application of the embodiments of this application in a real-world application scenario.

[0153] Next, taking a risk control modeling scenario based on vertical federated learning as an example, the training method of the federated learning model provided in this application embodiment will be explained. In the process of risk control modeling, to ensure the model's effectiveness, it is often necessary to obtain multi-dimensional and comprehensive user features to characterize the user profile. This often requires acquiring feature data from multiple sources. However, due to increasingly stringent user privacy protection, directly obtaining user data from data sources often becomes impractical. Vertical federated learning provides a solution for multi-party joint modeling without disclosing user privacy data, making it highly suitable for risk control modeling scenarios. See [link to relevant documentation] Figure 4 Vertical federated learning typically involves different participants jointly training a machine learning model. The labeled participant (usually only one) is called the active participant, while the unlabeled participant is called the passive participant. During model training, the participants only transmit encrypted model and gradient information, not other data, thus protecting user privacy.

[0154] In related technologies, see Figure 13 , Figure 13 This diagram illustrates a risk control model based on longitudinal federated learning. The modeling process includes: Step 301, the active device performs a brute-force attack based on the sample identifier and the retrospective month; Step 302, configure risk control modeling information; Step 303, execute risk control modeling; Step 304, score the risk control model; Step 305, write the risk control modeling report. Throughout the risk control modeling process, modelers need to configure a large amount of information, such as the table names, namespaces, and job IDs obtained from the brute-force attack. Ordinary modeling users need to carefully read the documentation and receive guidance from professionals to conduct federated modeling. Because the brute-force attack and modeling processes in federated modeling are time-consuming, user configuration errors can waste a significant amount of time, impacting both modeling efficiency and user experience.

[0155] The above analysis reveals that related federated learning systems expose a large number of system technical details unrelated to modeling to users, impacting user experience. Therefore, this application provides a training method for a federated learning model to address the problems of low modeling efficiency and poor user experience in risk control modeling based on vertical federated learning.

[0156] In practical implementation, for risk control modeling, the variables that users typically need to decide are only: the federated site number (site ID), the division of the training, validation, and test sets, the feature selection strategy, and the model parameters. All other configuration processes can be completed automatically by the system. See also Figure 14 , Figure 14 This application provides a training method for a risk control model based on federated learning, combined with... Figure 14 The steps shown are explained.

[0157] Step 401: The active device obtains the sample identifier and sends a sample confirmation request carrying the sample identifier to the passive device.

[0158] In practice, the active device sends a sample confirmation request carrying a sample identifier to the passive device, which can be seen as performing a credential stuffing attack. At this stage, the user needs to configure the sample file in the form of sample identifier (id), apply_month, y, x0, x1, ..., xn. Here, apply_month represents the time when the user's behavior attributes are sent. For example, in a risk control model for a user applying for a loan, apply_month could be the month the user applied for the loan. y represents the model label information, and {x0, x1, ..., xn} represents the feature data corresponding to the sample ID. This feature data is jointly provided by the active and passive devices participating in the federated model training.

[0159] In practice, the active party devices identify the active and passive party devices participating in the federated model training by specifying a federated site identifier (site ID). The system can automatically split the sample data according to the backtracking month (apply_month), and then perform brute-force attacks on the corresponding backtracking month for each split sample. After the brute-force attack is completed, the table name and namespace are automatically recorded. For months where the brute-force attack fails, multiple attempts can be made. If all attempts fail, manual processing is then required.

[0160] Step 402: The active device combines the first sample features provided locally that correspond to the sample identifier with the second sample features provided by the passive device to train a risk control model based on federated learning.

[0161] In practice, the information that must be determined by the user at this stage is the division of the training, validation, and test sets. The user only needs to specify which months will be used as the training set, and which months as the validation and test sets. Since the system records the table names and namespaces for each month, the system can automatically complete all configuration information once the user specifies the division of the training, validation, and test sets. Feature selection parameters and model parameters are optional; if the user does not configure them, the default configuration will be used.

[0162] Step 403: The active device scores each of the obtained risk control models to evaluate the predictive accuracy of the risk control models.

[0163] In actual implementation, all samples obtained from the brute-force attack are automatically scored. For samples with 'y', the effectiveness is evaluated, as are the effectiveness evaluations for the training set and validation set. If the test set has 'y', the effectiveness is also evaluated.

[0164] Step 404: The active device generates an evaluation report based on the scoring results of each risk control model.

[0165] In practice, users specify a report template, and the system automatically generates a modeling report based on the scoring and effect evaluation.

[0166] This application's embodiments streamline the process of vertical federated risk control modeling, identifying the variables that need to be determined by the user during the modeling process. This effectively avoids exposing too many details of the federated learning system to the modeling user, greatly simplifying the configuration process and improving the user experience; furthermore, it effectively reduces user errors, significantly improves the modeling efficiency of the federated learning system, and enhances the user experience for the active modeling user.

[0167] The following description continues to illustrate the exemplary structure of the training device 555 for the federated learning model provided in this application embodiment as a software module. In some embodiments, such as... Figure 2 As shown, the software modules stored in the training device 555 of the federated learning model in the memory 540 may include:

[0168] The acquisition module 5551 is used to acquire the sample identifier of the sample used to train the federated learning model, and the first sample feature of the sample stored locally.

[0169] The sending module 5552 is used to send a sample determination request carrying the sample identifier to at least one passive party device; wherein, the sample determination request is used by the passive party device to filter and obtain a second sample feature corresponding to the sample identifier from a local sample feature library;

[0170] The receiving module 5553 is used to receive response information sent by the passive device in response to the sample determination request;

[0171] The training module 5554 is used to train the federated learning model by combining the first sample feature and the second sample feature of the sample when the response information indicates that the second sample feature filtering corresponding to the sample identifier is completed.

[0172] In some embodiments, the acquisition module is further configured to acquire sample configuration information, the sample configuration information including sample feature range and sample identifier, the sample feature range being used to indicate the range to which the sample features of the samples used to train the federated learning model belong; determine the local sample features corresponding to the sample identifier in the local sample feature library according to the sample identifier; and filter the sample features that fall within the sample feature range according to the sample feature range as the first sample features.

[0173] In some embodiments, the acquisition module is further configured to acquire sample configuration information, the sample configuration information including a sample feature range; according to the sample feature range, filter sample features in the local sample feature library that fall within the sample feature range as first sample features; and use the sample identifier of the first sample feature as the sample identifier of the sample used to train the federated learning model. In some embodiments, the sample configuration information includes at least one passive party device identifier, and the sending module is further configured to send a sample determination request carrying the sample identifier to the corresponding passive party device according to the at least one passive party device identifier.

[0174] In some embodiments, the sending module is further configured to determine a plurality of candidate participant devices capable of co-training the federated learning model with the active party device; determine the priority of each candidate participant device; and, based on the priority, select at least one candidate participant device from the plurality of passive party devices as the passive party device.

[0175] In some embodiments, the training module is further configured to, when the first sample feature and the second sample feature belong to at least two feature ranges, combine the first sample feature and the second sample feature in each feature range to train a federated learning model to obtain a federated learning model corresponding to each feature range; evaluate the prediction accuracy of the federated learning model corresponding to each feature range to obtain an evaluation result; and output the evaluation result using a target output method.

[0176] In some embodiments, the sending module is further configured to, when the number of passive party devices is at least two, obtain the sending order corresponding to at least two passive party devices; according to the sending order, send a sample determination request carrying the sample identifier to each passive party device one by one; and, during the process of sending the sample determination request to each passive party device, stop sending the sample determination request when it is determined that the second sample feature corresponding to the sample identifier has been successfully filtered based on the response information returned by each passive party device.

[0177] In some embodiments, the training module is further configured to output sample determination failure information corresponding to the first sample identifier when the number of sample identifiers is at least two and the response information indicates that the second sample feature filtering corresponding to the first sample identifier among the at least two sample identifiers has failed, so as to prompt that the second sample feature filtering corresponding to the first sample identifier has failed.

[0178] In some embodiments, the training module is further configured to: determine the successfully filtered sample identifier as the second sample identifier when the response information indicates that the second sample feature corresponding to the sample identifier has been filtered and partially filtered successfully; filter the first sample feature corresponding to the second sample identifier from a plurality of first sample features corresponding to the sample identifier; and train the federated learning model by combining the first sample feature corresponding to the second sample identifier and the second sample feature corresponding to the second sample identifier.

[0179] It should be noted that the description of the apparatus in this application embodiment is similar to the description of the method embodiment above, and has similar beneficial effects as the method embodiment, so it will not be repeated.

[0180] This application provides a computer program product, including a computer program, characterized in that, when the computer program is executed by a processor, it implements the training method of the federated learning model provided in this application.

[0181] This application provides a computer-readable storage medium storing executable instructions. When these executable instructions are executed by a processor, they cause the processor to perform the method provided in this application, for example... Figure 3 The training method of the federated learning model is shown.

[0182] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0183] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0184] As an example, executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborating files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0185] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0186] In summary, the embodiments of this application can effectively avoid exposing too many details of the federated learning system to modeling users, greatly simplify the configuration process, and improve the user experience; and can effectively reduce user misoperation, effectively improve the modeling efficiency of the federated learning system, and enhance the user experience of active modeling users.

[0187] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. A training method for a federated learning model, characterized in that, Applied to an active-side device with a tag, the method includes: Obtain the sample identifier of the sample used to train the federated learning model, and the first sample feature of the sample stored locally; Send a sample determination request carrying the sample identifier to at least one passive party device; wherein, the sample determination request is used by the passive party device to filter and obtain a second sample feature corresponding to the sample identifier from a local sample feature library; Receive response information from the at least one passive device corresponding to the sample determination request; When the response information indicates that the second sample feature filtering corresponding to the sample identifier is completed, the federated learning model is trained by combining the first sample feature and the second sample feature of the sample. The step of obtaining the sample identifiers of the samples used to train the federated learning model, and the first sample features of the samples stored locally, includes: Obtain sample configuration information, which includes sample feature range and sample identifier; Based on the sample identifier, determine the local sample features corresponding to the sample identifier in the local sample feature library; Based on the range of sample features, sample features that fall within the range of sample features are selected as the first sample features.

2. The method according to claim 1, characterized in that, The sample configuration information includes at least one passive party device identifier, and sending a sample determination request carrying the sample identifier to at least one passive party device includes: Based on the at least one passive device identifier, a sample determination request carrying the sample identifier is sent to the corresponding passive device.

3. The method according to claim 1, characterized in that, Before sending the sample determination request carrying the sample identifier to at least one passive party device, the method further includes: Identify multiple candidate participant devices that can collaboratively train the federated learning model with the active participant device; Determine the priority of each of the candidate participant devices; Based on the priority, at least one candidate participant device is selected from the plurality of passive party devices as the passive party device.

4. The method according to claim 1, characterized in that, When the first sample feature and the second sample feature belong to at least two feature ranges, the step of training the federated learning model by combining the first sample feature and the second sample feature includes: By combining the first sample features and the second sample features within each of the aforementioned feature ranges, a federated learning model is trained to obtain a federated learning model corresponding to each of the aforementioned feature ranges. The method further includes: The prediction accuracy of the federated learning model corresponding to each of the aforementioned feature ranges is evaluated to obtain the evaluation results; The evaluation results are output using the target output method.

5. The method according to claim 1, characterized in that, When the number of passive party devices is at least two, sending a sample determination request carrying the sample identifier to at least one passive party device includes: Obtain the transmission order corresponding to at least two of the passive devices; According to the sending order, a sample determination request carrying the sample identifier is sent to the passive device one by one; During the process of sending the sample determination request to each of the passive party devices, when it is determined that the second sample feature corresponding to the sample identifier has been successfully filtered based on the response information returned by each of the passive party devices, the sending of the sample determination request is stopped.

6. The method according to claim 1, characterized in that, The method further includes: When the number of sample identifiers is at least two, and the response information indicates that the second sample feature filtering corresponding to the first sample identifier among the at least two sample identifiers has failed, the sample determination failure information corresponding to the first sample identifier is output to indicate that the second sample feature filtering corresponding to the first sample identifier has failed. When there are multiple sample identifiers, when the response information indicates that the second sample feature filtering corresponding to the sample identifier is completed, the federated learning model is trained by combining the first sample feature and the second sample feature of the sample, including: When the response information indicates that the second sample feature filtering corresponding to the sample identifier is completed and partially successful, the successfully filtered sample identifier is determined as the second sample identifier. From a plurality of first sample features corresponding to the sample identifier, filter the first sample features corresponding to the second sample identifier; The federated learning model is trained by combining the first sample feature corresponding to the second sample identifier and the second sample feature corresponding to the second sample identifier.

7. A training method for a federated learning model, characterized in that, Applied to an active-side device with a tag, the method includes: Obtain the sample identifier of the sample used to train the federated learning model, and the first sample feature of the sample stored locally; Send a sample determination request carrying the sample identifier to at least one passive party device; wherein, the sample determination request is used by the passive party device to filter and obtain a second sample feature corresponding to the sample identifier from a local sample feature library; Receive response information from the at least one passive device corresponding to the sample determination request; When the response information indicates that the second sample feature filtering corresponding to the sample identifier is completed, the federated learning model is trained by combining the first sample feature and the second sample feature of the sample. The step of obtaining the sample identifiers of the samples used to train the federated learning model, and the first sample features of the samples stored locally, includes: Obtain sample configuration information, which includes sample feature range; Based on the range of sample features, select sample features from the local sample feature library that fall within the range of sample features and use them as the first sample features. The sample identifier of the first sample feature is used as the sample identifier of the sample used to train the federated learning model.

8. A training device for a federated learning model, characterized in that, include: The acquisition module is used to acquire the sample identifier of the sample used to train the federated learning model, and the first sample feature of the sample stored locally. A sending module is configured to send a sample determination request carrying the sample identifier to at least one passive party device; wherein, the sample determination request is used by the passive party device to filter and obtain a second sample feature corresponding to the sample identifier from a local sample feature library; A receiving module is configured to receive response information sent by the at least one passive party device in response to the sample determination request; The training module is used to train the federated learning model by combining the first sample feature and the second sample feature of the sample when the response information indicates that the second sample feature filtering corresponding to the sample identifier is completed. The step of obtaining the sample identifiers of the samples used to train the federated learning model, and the first sample features of the samples stored locally, includes: Obtain sample configuration information, which includes sample feature range and sample identifier; Based on the sample identifier, determine the local sample features corresponding to the sample identifier in the local sample feature library; Based on the range of sample features, sample features that fall within the range of sample features are selected as the first sample features.

9. An electronic device, characterized in that, include: Memory, used to store executable instructions; A processor, when executing executable instructions stored in the memory, implements the training method of the federated learning model according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores executable instructions for implementing the training method of the federated learning model according to any one of claims 1 to 7 when executed by a processor.

11. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the training method of the federated learning model as described in any one of claims 1 to 7.

Images