Information classification method and device based on joint learning

Abstract

The present invention provides an information classification method and device based on joint learning. The method comprises: receiving data information uploaded by a participant through a server, wherein the data information at least comprises: the local model information of the participant and / or the parameters of the local model training corresponding to the local model information; screening the data information uploaded by the participant according to the preset conditions in the joint learning framework to obtain the classification of the data information; adding the classification of the data information to the classification task queue and sending the message to the participant; responding to the execution information of the local model uploaded by the participant, determining the training plan of the local model of the participant; encrypting the training plan and sending it to the participant. The present invention can classify the training data efficiently and in real time during model training to improve the training accuracy and stability of the local model of the participant.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to an information classification method and device based on joint learning. Background Art

[0002] As users become more aware of data sovereignty, how to ensure user data privacy has become the key to restricting machine learning algorithms, and classification is one of the most critical links in machine learning modeling. For example, in the credit investigation scenario, the quality of data uploaded by users is mixed into one training module, which will lead to low training accuracy.

[0003] How to efficiently and real-time classify training data during model training has become an urgent problem that needs to be solved in the current research field of artificial intelligence. Summary of the invention

[0004] In view of this, the embodiments of the present disclosure provide an information classification method and device based on joint learning to solve the problem that the prior art cannot classify model training data due to data silos.

[0005] A first aspect of the embodiments of the present disclosure provides an information classification method based on joint learning, including:

[0006] The server receives data information uploaded by the participant, wherein the data information at least includes: local model information of the participant and / or parameters of local model training corresponding to the local model information;

[0007] The data information uploaded by the participants is screened according to the preset conditions in the joint learning framework to obtain the classification of the data information;

[0008] Sending a message of adding the classification of the data information into the classification task queue to the participating party;

[0009] In response to the execution information of the local model uploaded by the participant, determine the training scheme of the local model of the participant;

[0010] The training plan is encrypted and sent to the participants.

[0011] A second aspect of the embodiments of the present disclosure provides an information classification method based on joint learning, including:

[0012] Responding to the key sent by the server, confirming as a participant in the federated learning architecture;

[0013] Data information uploaded to the server, wherein the data information at least includes: local model information of the participant and / or parameters of local model training corresponding to the local model information;

[0014] In response to a message from the server that the classification of the data information is added to a classification task queue;

[0015] Obtaining the encrypted training plan of the local model sent by the server;

[0016] The local model is trained and classified according to the training scheme, and the execution results after training and classification are uploaded to the server.

[0017] According to a third aspect of the embodiments of the present disclosure, there is provided an information classification device based on joint learning, comprising:

[0018] A receiving module, used for the server to receive data information uploaded by a participant, wherein the data information at least includes: local model information of image information and / or local model information of text information;

[0019] The preprocessing module is used to filter the data information uploaded by the participants according to the preset conditions in the joint learning framework to obtain the classification of the data information;

[0020] An allocation module, used for sending a message of adding the classification of the data information into the classification task queue to the participating parties;

[0021] A determination module, configured to respond to the execution information of the local model uploaded by the participant and determine the training scheme of the local model of the participant;

[0022] The sending module is used to encrypt the training plan and send it to the participants.

[0023] A fourth aspect of the embodiments of the present disclosure provides an information classification device based on joint learning, including:

[0024] A response module, used to respond to the key sent by the server and confirm as a participant in the joint learning architecture;

[0025] An uploading module, used for uploading data information to a server, wherein the data information at least includes: local model information of the participant and / or parameters of local model training corresponding to the local model information;

[0026] A receiving module, configured to respond to a message from the server to add the classification of the data information to a classification task queue;

[0027] An acquisition module is used to acquire the encrypted training plan of the local model sent by the server;

[0028] The training module is used to train and classify the local model according to the training scheme, and upload the execution results after training and classification to the server.

[0029] Compared with the prior art, the beneficial effects of the disclosed embodiment are as follows: receiving data information uploaded by the participant through the server, wherein the data information at least includes: the local model information of the participant and / or the parameters of the local model training corresponding to the local model information; filtering the data information uploaded by the participant according to the preset conditions in the joint learning framework to obtain the classification of the data information; sending the classification of the data information to the participant by adding the message of the classification task queue; responding to the execution information of the local model uploaded by the participant, determining the training plan of the local model of the participant; encrypting the training plan and sending it to the participant. The disclosed embodiment can efficiently and real-time classify the training data when conducting model training to improve the training accuracy and stability of the local model of the participant. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0031] Figure 1 is a schematic diagram of an application scenario of an embodiment of the present disclosure;

[0032] Figure 2 is a flow chart of an information classification method based on joint learning provided by an embodiment of the present disclosure;

[0033] Figure 3 is a flow chart of another information classification method based on joint learning provided by an embodiment of the present disclosure;

[0034] Figure 4 is a block diagram of another information classification device based on joint learning provided by an embodiment of the present disclosure;

[0035] Figure 5 is a block diagram of an information classification device based on joint learning provided by an embodiment of the present disclosure;

[0036] Figure 6 It is a schematic diagram of a computer device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0037] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present disclosure. However, it should be clear to those skilled in the art that the present disclosure may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present disclosure with unnecessary details.

[0038] Federated learning refers to the comprehensive use of various AI (Artificial Intelligence) technologies to jointly explore the value of data and give rise to new intelligent formats and models based on joint modeling, while ensuring data security and user privacy. Federated learning has at least the following characteristics:

[0039] (1) A weakly centralized joint training model in which participating nodes control their own data to ensure data privacy and security in the process of co-creating intelligence.

[0040] (2) In different application scenarios, we use screening and / or combination of AI algorithms and privacy-preserving computing to establish multiple model aggregation optimization strategies to obtain high-level, high-quality models.

[0041] (3) Under the premise of ensuring data security and user privacy, methods to improve the performance of the federated learning engine are obtained based on multiple model aggregation optimization strategies. The performance method can be achieved by solving problems including computing architecture parallelism, information interaction in large-scale cross-domain networks, intelligent perception, and exception handling mechanisms to improve the overall performance of the federated learning engine.

[0042] (4) Obtain the needs of multiple users in each scenario, determine a reasonable assessment of the true contribution of each joint participant through a mutual trust mechanism, and distribute incentives.

[0043] Based on the above methods, we can establish an AI technology ecosystem based on federated learning, give full play to the value of industry data, and promote the implementation of scenarios in vertical fields.

[0044] A method and device for information classification based on joint learning according to an embodiment of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0045] Figure 1 Schematic diagram of a joint learning architecture of an embodiment of the present disclosure. Figure 1 As shown, the architecture of the federated learning may include a server (central node) 101 and participants 102 , 103 , and 104 .

[0046] In the joint learning process, the basic model can be established by the server 101, and the server 101 sends the model to the participants 102, 103, and 104 with which the communication connection is established. The basic model can also be uploaded to the server 101 by any participant after establishment, and the server 101 sends the model to other participants with which the communication connection is established. The participants 102, 103, and 104 build the model according to the downloaded basic structure and model parameters, use local data to train the model, obtain updated model parameters, and encrypt and upload the updated model parameters to the server 101. The server 101 aggregates the model parameters sent by the participants 102, 103, and 104 to obtain the global model parameters, and transmits the global model parameters back to the participants 102, 103, and 104. The participants 102, 103, and 104 iterate their respective models according to the received global model parameters until the model finally converges, thereby realizing the training of the model. In the joint learning process, the data uploaded by participants 102, 103 and 104 are model parameters, local data will not be uploaded to server 101, and all participants can share the final model parameters, so joint modeling can be achieved on the basis of ensuring data privacy. It should be noted that the number of participants is not limited to the three as described above, but can be set as needed, and the embodiments of the present disclosure do not limit this.

[0047] Figure 2 It is a flowchart of an information classification method based on joint learning provided by an embodiment of the present disclosure. Figure 2 The information classification method based on joint learning can be obtained by Figure 1 The server executes. Figure 2 As shown, the information classification method based on joint learning includes:

[0048] S201, the server receives data information uploaded by the participants, wherein the data information at least includes: the local model information of the participants and / or the parameters of the local model training corresponding to the local model information; the local model information may include the target application environment value, expected value and hardware coefficient of the device for the local model training, etc., which is not limited by the present invention, and the server may receive the data in real time according to the needs of the participants.

[0049] Specifically, the server initializes the data information (such as the relevant data information of the training model such as the image classification for judging whether the gas has a flue or not) and selects the qualified participants (qualified means that the data meets the requirements and the hardware meets the requirements); the server can receive the classification information uploaded by the participants in the following ways:

[0050] Step 1: Confirm that the party sending the data information has joined the joint learning framework;

[0051] Step 2: Initialize the data information sent by the participants according to the preset conditions;

[0052] Step 3: Encrypt the initialized data information and send the encryption key to the network structure corresponding to the participants joining the federated learning architecture;

[0053] The initialized data information at least includes: basic information of the participants and the participants' satisfaction with the data.

[0054] S202, filtering the data information uploaded by the participants according to the preset conditions in the joint learning framework to obtain a classification of the data information.

[0055] Specifically, based on the joint learning framework, when receiving data information uploaded by a participant, the preset conditions corresponding to the participant are taken. Since one server can correspond to multiple participants, when a participant is determined to join the joint learning framework, a corresponding label can be set for the participant based on the information actively provided by the participant. Then, appropriate preset conditions are set for the participant without the need for the participant to upload local data, ensuring the security of the user's data privacy.

[0056] S203, sending a message of adding the classification of data information into the classification task queue to the participating parties.

[0057] Specifically, the server may classify the data information uploaded by the participants according to the preset conditions, and send the classification results to the corresponding classification task queues, wherein each participant is assigned a classification task queue.

[0058] S204, responding to the execution information of the local model uploaded by the participant, and determining the training plan of the local model of the participant.

[0059] Specifically, when the participant receives the uploaded data information that has entered the classification task queue of the server, the participant will upload the execution information of the local model to the server. At this time, the server will respond to the execution information of the local model uploaded by the participant and determine the training plan of the participant's local model for the participant.

[0060] Furthermore, the training scheme for determining the local model of the participant can be implemented in the following manner: Step 1, responding to the execution information of the local model uploaded by the participant;

[0061] Step 2: Match the execution information with the classification of the data information;

[0062] When the matched information is the local model sent by the updating participant, the target number of training rounds for the local model to be trained by the participant is calculated, and the local model is used as the updating training model scheme;

[0063] When the matching information is that the participant is currently training the local model for the last round, the average value of the parameters of the participant's last round of local model training is calculated, and the average value is used as the final training model parameter, and the local model is used as the training model solution.

[0064] Among them, using the average value as the final training model parameter can be achieved in the following way: first, verify the average value of the parameters of the last round of local model training of the computing participant; then, when the verification result is within the preset range, it is sent to the participant as the final training model parameter.

[0065] Among them, calculating the value of the target number of training rounds for the participant to train the local model can be achieved in the following way: it can be achieved by judging whether the number of local model training rounds of the current participant meets the preset convergence value; if so, determining the number of training rounds; if not, feeding back to the participant to continue updating the local model until the loss function convergence value is met.

[0066] S205, encrypt the training plan and send it to the participants.

[0067] According to the technical solution provided by the embodiment of the present disclosure, the server receives data information uploaded by the participant, wherein the data information at least includes: the local model information of the participant and / or the parameters of the local model training corresponding to the local model information; screens the data information uploaded by the participant according to the preset conditions in the joint learning framework to obtain the classification of the data information; sends the classification of the data information to the participant by adding the message of the classification task queue; responds to the execution information of the local model uploaded by the participant, determines the training plan of the local model of the participant; encrypts the training plan and sends it to the participant. In addition, the training data can be efficiently and real-time classified during model training to improve the training accuracy and stability of the local model of the participant.

[0068] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, which will not be described one by one here.

[0069] Figure 3 It is a flowchart of an information classification method based on joint learning provided by an embodiment of the present disclosure. Figure 3 The information classification method based on joint learning can be obtained by Figure 1 Executed by the participants. Figure 3 As shown, the information classification method based on joint learning includes:

[0070] S301, in response to the key sent by the server, confirming as a participant in the joint learning architecture.

[0071] Participants need to obtain authentication from the server before joining the federated learning framework, that is, select qualified participants (qualified participants can be data standards or hardware standards that meet service preset conditions, etc.).

[0072] S302, uploading data information to the server, wherein the data information at least includes: local model information of the participant and / or parameters of local model training corresponding to the local model information.

[0073] S303, responding to the message that the server adds the classification of data information into the classification task queue.

[0074] S304, obtaining the encrypted local model training plan sent by the server.

[0075] Specifically, when the participant knows that the uploaded data information has entered the classification task queue in the server, it will upload the execution information of the current local model to the server; the server will send the relevant training model solution to the participant through the classification task queue based on the execution information.

[0076] S305: Train and classify the local model according to the training plan, and upload the execution results after training and classification to the server.

[0077] Specifically, training and classifying the local model according to the training scheme and uploading the execution results after training and classification to the server can be achieved in the following ways:

[0078] When the training scheme is a training model update scheme, the parameters of the local model of the participant are updated using the neural network algorithm of joint learning, and the information of the updated local model is uploaded to the server;

[0079] When the training plan is a training model plan, obtain the training model parameters sent by the server to perform the last round of model training on the local model, and upload the last round of model training results to the server.

[0080] Among them, updating the parameters of the local model of the participants using the neural network algorithm of joint learning can be achieved in the following ways: obtaining the data of the local training model; preprocessing the data of the local training model; and performing structured integration based on the preprocessing results.

[0081] According to the technical solution provided by the embodiment of the present disclosure, the participant confirms that it is a participant in the joint learning architecture by responding to the key issued by the server; uploads data information to the server, wherein the data information at least includes: the local model information of the participant and / or the parameters of the local model training corresponding to the local model information; responds to the message of the server adding the classification of the data information to the classification task queue;

[0082] Obtain the encrypted local model training plan sent by the server; classify the local model according to the training plan, and upload the execution results after training classification to the server. In addition, the training accuracy and stability of the local model of the participant can be improved by efficiently and real-time classification of training data during model training.

[0083] The following are embodiments of the device disclosed herein, which can be used to execute the method embodiments disclosed herein. For details not disclosed in the device embodiments disclosed herein, please refer to the method embodiments disclosed herein.

[0084] Figure 4 It is a schematic diagram of an information classification device based on joint learning provided in an embodiment of the present disclosure.

[0085] like Figure 4 As shown, the information classification device based on joint learning includes:

[0086] The receiving module 401 is configured to receive data information uploaded by a participant on the server, wherein the data information at least includes: local model information of the participant and / or parameters of local model training corresponding to the local model information;

[0087] The pre-processing module 402 is configured to filter the data information uploaded by the participants according to the preset conditions in the joint learning framework to obtain the classification of the data information;

[0088] The allocation module 403 is configured to send a message to a participant for adding the classification of data information to the classification task queue;

[0089] The determination module 404 is configured to respond to the execution information of the local model uploaded by the participant and determine the training scheme of the local model of the participant;

[0090] The sending module 405 is configured to encrypt the training plan and send it to the participants.

[0091] According to the technical solution provided by the embodiment of the present disclosure, the server receives data information uploaded by the participant, wherein the data information at least includes: the local model information of the participant and / or the parameters of the local model training corresponding to the local model information; screens the data information uploaded by the participant according to the preset conditions in the joint learning framework to obtain the classification of the data information; sends the classification of the data information to the participant by adding the message of the classification task queue; responds to the execution information of the local model uploaded by the participant, determines the training plan of the local model of the participant; encrypts the training plan and sends it to the participant. In addition, the training data can be efficiently and real-time classified during model training to improve the training accuracy and stability of the local model of the participant.

[0092] Figure 5 It is a schematic diagram of an information classification device based on joint learning provided in an embodiment of the present disclosure.

[0093] like Figure 5 As shown, the information classification device based on joint learning includes:

[0094] A response module 501 is configured to respond to a key sent by the server and confirm as a participant in the joint learning architecture;

[0095] The uploading module 502 is configured to upload data information to the server, wherein the data information at least includes: local model information of image information and / or local model information of text information;

[0096] A receiving module 503 is configured to respond to a message from the server that the classification of the data information is added to a classification task queue;

[0097] An acquisition module 504 is configured to acquire the encrypted training scheme of the local model sent by the server;

[0098] The training module 505 is configured to perform training classification on the local model according to the training scheme, and upload the execution result after training classification to the server.

[0099] According to the technical solution provided by the embodiment of the present disclosure, the participant confirms that it is a participant in the joint learning architecture by responding to the key issued by the server; uploads data information to the server, wherein the data information at least includes: the local model information of the participant and / or the parameters of the local model training corresponding to the local model information; responds to the message of the server adding the classification of the data information to the classification task queue;

[0100] Obtain the encrypted local model training plan sent by the server; classify the local model according to the training plan, and upload the execution results after training classification to the server. In addition, the training accuracy and stability of the local model of the participant can be improved by efficiently and real-time classification of training data during model training.

[0101] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present disclosure.

[0102] Figure 6 Schematic diagram of a computer device 6 provided in an embodiment of the present disclosure. Figure 6As shown, the computer device 6 of this embodiment includes: a processor 601, a memory 602, and a computer program 603 stored in the memory 602 and executable on the processor 601. When the processor 601 executes the computer program 603, the steps in the above-mentioned various method embodiments are implemented. Alternatively, when the processor 601 executes the computer program 603, the functions of the modules / units in the above-mentioned various device embodiments are implemented.

[0103] Exemplarily, the computer program 603 may be divided into one or more modules / units, which are stored in the memory 602 and executed by the processor 601 to complete the present disclosure. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program 603 in the computer device 6.

[0104] The computer device 6 may be a desktop computer, a notebook computer, a PDA, a cloud server, or other computer devices. The computer device 6 may include but is not limited to a processor 601 and a memory 602. Those skilled in the art will appreciate that Figure 6 It is only an example of computer device 6 and does not constitute a limitation of computer device 6. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device may also include input and output devices, network access devices, buses, etc.

[0105] The processor 601 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0106] The memory 602 may be an internal storage unit of the computer device 6, for example, a hard disk or memory of the computer device 6. The memory 602 may also be an external storage device of the computer device 6, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device 6. Further, the memory 602 may also include both an internal storage unit and an external storage device of the computer device 6. The memory 602 is used to store computer programs and other programs and data required by the computer device. The memory 602 may also be used to temporarily store data that has been output or is to be output.

[0107] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0108] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0109] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this disclosure.

[0110] In the embodiments provided in the present disclosure, it should be understood that the disclosed apparatus / computer equipment and methods can be implemented in other ways. For example, the apparatus / computer equipment embodiments described above are merely schematic. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation. Multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection of the apparatus or unit, which may be electrical, mechanical or other forms.

[0111] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0112] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0113] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present disclosure implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. The computer program may include computer program code, and the computer program code may be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electric carrier signals and telecommunication signals.

[0114] The above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should all be included in the protection scope of the present disclosure.

Claims

1. An information classification method based on joint learning, characterized in that: include: The server receives data information uploaded by the participant, wherein the data information at least includes: local model information of the participant's image information and / or local model information of the text information and parameters of local model training corresponding to the local model information, wherein the local model information includes at least one of a target application environment value, an expected value, and a hardware coefficient of a device for local model training; The data information uploaded by the participants is screened according to the preset conditions in the joint learning framework to obtain the classification of the data information; including: setting a corresponding label for the participant according to the data information sent by the participant; Sending a message of adding the classification of the data information into the classification task queue to the participating party; In response to the execution information of the local model uploaded by the participant, determine the training scheme of the local model of the participant; Encrypting the training plan and sending it to the participants; Determining the training scheme of the local model of the participant includes: Respond to the execution information of the local model uploaded by the participant; matching the execution information with the classification of the data information; When the matched information is the local model sent by the updating participant, the target number of training rounds for the local model of the participant is calculated, and the training scheme of the local model is determined to be the updating training model scheme; When the matching information is that the participant is currently training the local model for the last round, the average value of the parameters of the participant's last round of local model training is calculated, and the average value is used as the final training model parameter to determine the training scheme of the local model as the training model scheme.

2. The method according to claim 1, characterized in that The data information uploaded by the participants received by the server includes: Confirm that the party sending the data information has joined the federated learning framework; Initialize the data information sent by the participants according to the preset conditions; Encrypt the initialized data information and send the encryption key to the network structure corresponding to the participants joining the federated learning architecture; The initialized data information at least includes: the basic information of the participants and the satisfaction of the participants with the data.

3. The method according to claim 1, characterized in that Taking the average value as the final training model parameter includes: Calculating the average value of the parameters of the last round of local model training of the participant for verification; When the verification result is within the preset range, it is sent to the participant as the final training model parameter.

4. The method according to claim 1, characterized in that: The value for calculating the target number of training rounds for the participant to train the local model includes: Determine whether the number of local model training rounds of the current participant meets the preset convergence value; If satisfied, determine the number of training rounds; If not, the participant is fed back to continue updating the local model until the loss function converges to a certain value.

5. An information classification method based on joint learning, characterized in that: include: Responding to the key sent by the server, confirming as a participant in the federated learning architecture; Data information uploaded to a server, so that the server sets a corresponding label for the participant according to the data information sent by the participant, and screens the data information uploaded by the participant according to the preset conditions in the joint learning framework to obtain the classification of the data information; wherein the data information at least includes: local model information of the image information and / or local model information of the text information of the participant and parameters of local model training corresponding to the local model information, and the local model information includes at least one of the target application environment value, expected value and hardware coefficient of the device for local model training; In response to the message that the server adds the classification of the data information to the classification task queue, the execution information of the local model is uploaded to the server so that the server matches the execution information with the classification of the data information. When the matching information is the local model sent by the updating participant, the value of the target number of training rounds for the local model of the participant is calculated, and the local model is determined as the updated training model scheme. When the matching information is the last round of training for the local model of the participant, the average value of the parameters of the last round of local model training of the participant is calculated, and the average value is used as the final training model parameter, the local model is determined as the training model scheme, and the training scheme is encrypted and sent to the participant; Obtaining the encrypted training plan of the local model sent by the server; The local model is trained and classified according to the training scheme, and the execution results after training and classification are uploaded to the server.

6. The method according to claim 5, characterized in that Training and classifying the local model according to the training scheme, and uploading the execution results after training and classification to the server include: When the training scheme is a training model update scheme, the parameters of the local model of the participant are updated using the neural network algorithm of joint learning, and the information of the updated local model is uploaded to the server; When the training scheme is a training model scheme, the training model parameters sent by the server are obtained to perform a final round of model training on the local model, and the final round of model training results are uploaded to the server.

7. The method according to claim 6, characterized in that Updating the parameters of the local model of the participant using the neural network algorithm of joint learning includes: Get data for local training models; Preprocess the data for local training models; Perform structured integration based on preprocessing results.

8. An information classification device based on joint learning, characterized in that: include: A receiving module, configured to receive data information uploaded by a participant on a server, wherein the data information at least includes: local model information of the participant's image information and / or local model information of the text information and parameters of local model training corresponding to the local model information, wherein the local model information includes at least one of a target application environment value, an expected value, and a hardware coefficient of a device for local model training; The preprocessing module is used to filter the data information uploaded by the participants according to the preset conditions in the joint learning framework to obtain the classification of the data information; including: setting a corresponding label for the participant according to the data information sent by the participant; An allocation module, used for sending a message of adding the classification of the data information into the classification task queue to the participating parties; A determination module, configured to respond to the execution information of the local model uploaded by the participant and determine the training scheme of the local model of the participant; A sending module, used for encrypting the training plan and sending it to the participant; The determination module is specifically used to: respond to the execution information of the local model uploaded by the participant; match the execution information with the classification of the data information; when the matched information is the local model sent by the updated participant, calculate the value of the target number of training rounds for the local model trained by the participant, and determine that the training scheme of the local model is the updated training model scheme; when the matched information is that the participant is currently training the local model for the last round, calculate the average value of the parameters of the last round of local model training of the participant, and use the average value as the final training model parameter, and determine that the training scheme of the local model is the training model scheme.

9. An information classification device based on joint learning, characterized in that: include: A response module, used to respond to the key sent by the server and confirm as a participant in the joint learning architecture; An upload module, used for uploading data information to a server, so that the server sets a corresponding label for the participant according to the data information sent by the participant, and screens the data information uploaded by the participant according to the preset conditions in the joint learning framework to obtain a classification of the data information; wherein the data information at least includes: local model information of the participant's image information and / or local model information of the text information and parameters of local model training corresponding to the local model information, and the local model information includes at least one of the target application environment value, expected value and hardware coefficient of the device for local model training; A receiving module, configured to respond to a message from the server that the classification of the data information is added to a classification task queue, upload execution information of the local model to the server, so that the server matches the execution information with the classification of the data information, and when the matched information is the local model sent by the updating participant, calculate the value of the target number of training rounds for the local model trained by the participant, and determine the local model as an updated training model scheme; when the matched information is the last round of training for the local model currently trained by the participant, calculate the average value of the parameters of the last round of local model training of the participant, and use the average value as the final training model parameter, determine the local model as a training model scheme, and encrypt the training scheme and send it to the participant; An acquisition module is used to acquire the encrypted training plan of the local model sent by the server; The training module is used to train and classify the local model according to the training scheme, and upload the execution results after training and classification to the server.

Images