Information processing device, information processing system, information processing method, and program

Abstract

To provide a device, system, method, and program that can allow for flexible changes to the scope of permitted data use, reduce the risk of data use between competing organizations, and enable safer and more effective data management.SOLUTION: An information processing device is used for integrating local models through federated learning to build a global model, and includes registration means for registering a local model trained by machine learning using a dataset held by the information processing device and reception means for receiving a specification of the scope of permitted use of the registered local model.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing system, an information processing method, and a program used to build a global model by integrating local models through federated learning. [Background technology]

[0002] Federated learning is a machine learning technique that allows learning to occur while keeping data distributed rather than aggregated. It is a technology that enables the construction of models that utilize training data between multiple clients, as if the data were linked, while ensuring privacy and security.

[0003] As an example of such federated learning, a technique is known in which each node learns a local model, and the server updates the data of the global model based on the data of each local model acquired from each node (see, for example, Patent Document 1). Summary of the Invention [Problem to be solved by the invention]

[0004] Previously, federated learning only allowed clients to choose whether to participate in federated learning, and once data was registered, all other clients participating in the federated learning could use that data. As a result, the scope of data federation was uniformly determined, which created the problem of not being able to flexibly change the scope of data usage permission to suit the security requirements and business needs of each organization.

[0005] Additionally, some of the organizations participating in federated learning are competitors, and there is a problem in that it is not possible to guarantee the safe use of data between competing organizations.

[0006] Therefore, it is desirable to provide an apparatus, system, method, and program that can flexibly change the scope of permitted data use, reduce the risk of data use between competing clients, and enable safer and more effective data management. [Means for solving the problem]

[0007] In view of the above-described problems, the present invention provides an information processing device used to construct a global model by integrating local models through federated learning, comprising: a registration means for registering data of a local model obtained by machine learning using a dataset held by the information processing device; a receiving means for receiving a specification of the range of permitted use of the registered data; An information processing device is provided, including: [Effects of the Invention]

[0008] According to the present invention, it is possible to provide an apparatus, system, method and program that allows the scope of permitted data use to be flexibly changed, reduces the risk of data use between competing organizations, and enables safer and more effective data management. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram showing an example of the configuration of an information processing system according to an embodiment of the present invention. [Figure 2] FIG. 1 is a diagram showing an example of a hardware configuration of an information processing device that constitutes an information processing system. [Figure 3] FIG. 2 is a block diagram showing an example of the functional configuration of an information processing device and a management device. [Figure 4] 10 is a flowchart showing an example of processing until a client participates in federated learning. [Figure 5] FIG. 10 is a diagram showing an example of a model use permission setting screen. [Figure 6] FIG. 10 is a diagram showing an example of a screen for selecting a coalition partner. [Figure 7] FIG. 10 is a diagram showing an example of a screen for applying to use a model. [Figure 8] FIG. 10 is a diagram showing an example of a screen for approving or denying a request to use a model. [Figure 9] 10 is a flowchart showing an example of a process for constructing a model by federated learning. DETAILED DESCRIPTION OF THE INVENTION

[0010] 1 is a diagram showing an example of the configuration of an information processing system according to this embodiment. The information processing system is a system that performs federated learning and includes client devices (hereinafter abbreviated as clients) 10 as multiple information processing devices, and a central server (hereinafter abbreviated as central) 11 as a management device.

[0011] The central 11 is connected to a storage device 12 that stores data to be held by the central 11 (central data, central model, etc.), and is connected to each client 10 via a network 13. The central model is a global model constructed by federated learning.

[0012] Each client 10 is installed, for example, at each user or organization. For example, client A is installed at organization A, client B is installed at organization B different from organization A, and client N is installed at organization N different from organizations A and B. A storage device 14 that stores data (client data, client model, etc.) to be held by each client 10 is connected to each client 10. The client model is a local model on which each client 10 performs machine learning using the client data as a dataset.

[0013] The central 11 generates an initial model as one of the central models and distributes the initial model to each client 10. Models such as the initial model, central model, and client model are neural networks, machine learning models, or the like.

[0014] A neural network includes multiple processing layers, each of which contains multiple nodes or neurons. The processing layers include an input layer, one or more hidden layers, and an output layer. Each node or neuron is assigned a weight, which is adjusted during the learning process.

[0015] A machine learning model is an algorithm that, when given some input, evaluates the content of the input and outputs the evaluation result as an output value. The algorithm is, for example, a function, and the simplest one is a linear function. The function includes a coefficient that is adjusted in the learning process as one of its parameters. These models are trained using training data in the form of pairs of example questions and correct answers, and the parameters are adjusted.

[0016] Each client 10 performs machine learning on the distributed initial model using the client data it holds, and constructs a client model. Each client 10 transmits the data of the constructed client model to the central 11. The data of the client model includes the above-mentioned parameters, the output of the client model, etc.

[0017] The central 11 integrates each client model using data acquired from each client 10, creates a new central model, updates the central model, and constructs the central model. The constructed central model can be used for each of the clients 10 that have participated in federated learning as federated partners for their own tasks, business, etc.

[0018] In order to perform federated learning, the information processing system may include, in addition to clients 10 that participate in federated learning, information processing devices 15 that do not participate in the federated learning. The information processing devices 15 are connected to a central 11 via a network 16, and are connected to a storage device 17 that stores data to be held by the information processing devices 15. The central 11 may distribute a central model constructed by federated learning to information processing devices 15 that are not participating in the federated learning.

[0019] The networks 13 and 16 may be wired networks or wireless networks such as wireless LANs (Local Area Networks). The information processing system is not limited to the multiple clients 10, the central 11, and the information processing devices 15, and may include other devices.

[0020] In the example shown in Figure 1, the central 11 and the storage device 12, and the client 10 and the storage device 14 are each separate devices, but this is not limited to this, and the central 11 may be equipped with the storage device 12, and the client 10 may be equipped with the storage device 14.

[0021] 2 is a diagram showing an example of the hardware configuration of the client 10 as an information processing device. Note that the hardware configurations of the central 11 and the information processing device 15 are similar to those of the client 10, and therefore descriptions of the hardware configurations of the central 11 and the information processing device 15 will be omitted.

[0022] The client 10 is constructed by a computer and includes a CPU (Central Processing Unit) 20, a ROM (Read Only Memory) 21, a RAM (Random Access Memory) 22, a HD (Hard Disk) 23, and a HDD (Hard Disk Drive) controller 24. The client 10 also includes a display 25, an external device connection I / F 26, a network I / F 27, a bus line 28, a keyboard 29, a pointing device 30, a DVD-RW (Digital Versatile Disk Rewritable) drive 31, and a media I / F 32.

[0023] The CPU 20 controls the overall operation of the client 10. The ROM 21 stores programs used to drive the CPU 20, such as an IPL (Initial Program Loader), which is first read into the RAM 22 and executed when the client 10 is started up. The RAM 22 provides a working area for the CPU 20. The HD 23 stores various data such as programs. The HD 23 may be used as the storage device 14 connected to the client 10. The HDD controller 24 controls the reading and writing of various data from and to the HD 23 under the control of the CPU 20. Note that the client 10 may include a storage device other than the HD 23, such as an SSD (Solid State Drive).

[0024] The display 25 displays various types of information such as a cursor, menus, windows, characters, or images. The external device connection I / F 26 is an interface for connecting various types of external devices. In this case, the external devices are, for example, USB (Universal Serial Bus) memories or printers. The network I / F 27 is an interface for data communication using the networks 13 and 16. The bus line 28 is an address bus, a data bus, or the like for electrically connecting each component such as the CPU 20.

[0025] The keyboard 29 is a type of input means equipped with multiple keys for inputting characters, numbers, various instructions, etc. The pointing device 30 is a type of input means for selecting and executing various instructions, selecting a processing target, moving a cursor, etc. The DVD-RW drive 31 controls reading and writing of various data from a DVD-RW 33, which is an example of a removable recording medium. Note that the recording medium is not limited to the DVD-RW 33, but may be a DVD-R, CD-RW, CD-R, etc. The media I / F 32 controls reading and writing of data from a recording medium 34, such as a flash memory.

[0026] The client 10 may also include other hardware such as a sound input device such as a microphone, a sound output device such as a speaker, and an imaging device such as a camera.

[0027] The client 10, central 11, and information processing device 15 are not limited to computers such as general PCs shown in FIG. 2, but may be devices such as MFPs (Multi-Function Peripherals) that can perform federated learning.

[0028] 3 is a block diagram showing an example of the functional configuration of client 10 and central 11. Client 10 and central 11 cooperate to perform federated learning, and have multiple functions to realize this processing, and each function can be realized by one or more processing circuits. Here, the processing circuit includes a processor programmed to execute each function by software, such as a processor implemented by an electronic circuit, as well as devices such as an ASIC (Application Specific Integrated Circuit), DSP (Digital Signal Processor), FPGA (Field Programmable Gate Array), and conventional circuit modules designed to execute each of the above functions.

[0029] The client 10 includes a transmitting unit 40, a receiving unit 41, a registration unit 42, a reception unit 43, and a learning unit 44. The transmitting unit 40 transmits client model data and the like to the central 11. The receiving unit 41 receives an initial model (initial model data) distributed from the central 11, a central model (central model data) created by integrating the received client models, and the like. The receiving unit 41 functions as an acquiring unit and acquires related information and the like from the central 11.

[0030] The registration unit 42 transmits data of the client model obtained by machine learning to the central 11 via the transmission unit 40, and registers the data in the storage device 12 connected to the central 11. In detail, the registration unit 42 requests the central 11 to register the data transmitted to the storage device 12, and the central 11 registers the data by saving the data in the storage device 12. The registration unit 42 also transmits related information relating to the client model constructed by the client 10 to the central 11, and registers the information in association with the relevant data. The registration unit 42 may also register the information in the storage device 14 connected to the client 10.

[0031] The related information is, but is not limited to, information such as the industry, the business that utilizes the relevant data, and the intended use. The industry includes retail, wholesale, food, communications, agriculture, forestry, mining, construction, transportation, finance, insurance, medical care, welfare, and services. The business includes research, development, production, sales, inquiries, surveys, and daily reports. The intended use is the purpose for which the model is used, and includes predicting stock prices and energy demand, and detecting objects.

[0032] The reception unit 43 displays a screen for a user or organization using the client 10 and receives input from the user, etc. The organization may be a company, department, group, etc. The screen may be, for example, a related information input screen, a security policy setting screen for registered data, a model selection screen, an alliance partner selection screen, etc. The security policy is a rule for maintaining the security of registered data, and can be set as the range of permitted data use, access level, etc. The range of permitted data use is, for example, the range of users, etc. The access level is the level of authority to access data assigned to users, etc., and may include a level that allows access to all data or a level that restricts access to part of the data, etc.

[0033] The learning unit 44 performs machine learning on the initial model distributed from the central 11, using the client data held by the client 10 as learning data. The machine learning method is a generally well-known method, and therefore a description thereof will be omitted here.

[0034] The central 11 includes a transmitting unit 50, a receiving unit 51, a registration processing unit 52, an update unit 53, and a judgment unit 54. The transmitting unit 50 functions as a distribution unit, and transmits and distributes the initial model, the central model, etc. to each client 10. The receiving unit 51 receives and acquires data of the client model that has been machine-learned by each client 10 from each client 10.

[0035] The registration processing unit 52 stores and registers the client model acquired from each client 10 in the storage device 12. The registration processing unit 52 also stores the updated central model in the storage device 12.

[0036] The update unit 53 integrates the stored client models and updates the central model. The central model is updated by creating a new central model using the client models and client model outputs acquired from each client 10. The client models can be integrated, for example, by averaging the above parameters of each client model.

[0037] The judgment unit 54 determines whether an update termination condition is satisfied. The update termination condition may be, for example, whether the update of the central model has been repeated a predetermined number of times, whether a convergence condition has been satisfied, or the like. The convergence condition may be, for example, whether the accuracy rate or recall rate is equal to or greater than a threshold. Note that these are merely examples and are not limited to these. The accuracy rate indicates how accurate the prediction or classification by the created central model is, and the recall rate indicates, when the correct answer is A, how well the prediction by the created central model was A, or how well the classification by the central model was A.

[0038] The more times the model is trained, the more detailed the data is learned, and the more repetitions are made, the more detailed the prediction or classification will be taken into account, making it difficult to make accurate predictions or classifications. This indicates a state called overfitting, and the update termination condition can be set as a condition that stops the update before the overfitting state is reached.

[0039] 4 is a flowchart showing an example of the processing up to when the client 10 participates in federated learning. In step 100, a user uses the client 10 to establish communication with the central 11. The user may be an individual who uses the client 10, an employee of a company, or the like.

[0040] In step 101, the client 10 transmits data for creating a model to the central 11 and registers the data. The data for creating a model includes the client model and the output of the client model. The user inputs the necessary information into the screen displayed on the client 10 and transmits it together with the above data to the central 11. The necessary information is information related to the client model, such as the industry, the business in which the relevant data will be used, and the purpose of use.

[0041] In step 102, the client 10 displays a screen for setting a security policy for the registered model and accepts the user's setting of the security policy. The client 10 accepts specifications such as the scope of permitted use of the model and the access level as the security policy. The client 10 transmits the set security policy to the central 11, and the central 11 stores it in association with the registered data.

[0042] In step 103, the client 10 displays the registered models and accepts a user's selection. The user can refer to information about tasks related to the registered models and select a model according to the task they wish to perform. A task represents an operation performed by a model, such as classification, regression, or clustering. In addition to accepting the model selection, the client 10 can also accept parameters related to the model and set the accepted parameters as initial values. The client 10 transmits information about the selected model to the central 11 and registers it as a model the user wishes to use. At this time, the information about the selected model can be saved in association with the above data and related information.

[0043] In step 104, the client 10 displays a screen for selecting a federation partner and accepts the user's selection of a federation partner. A federation partner is another client that will participate in federated learning. The screen for selecting a federation partner includes information such as available datasets, the industries of participating clients 10, and data quality, and the user can select a federation partner based on this information.

[0044] This selection allows for efficient and purposeful federated learning. In addition to the user's selection, the client 10 may automatically select the most suitable federated partner based on the data characteristics of the client 10, the selected model, the set security policy, etc.

[0045] The client 10 sends information about the selected federated partner to the central 11, permits the use of the registered data, and registers the partner as a partner for federated learning. At this time, the information about the selected federated partner can be saved in association with the above data and related information.

[0046] In step 105, client 10 participates in the federated learning process, working with federated partners to jointly train models, and then proceeds to step 106, completing the process up to which client 10 participates in federated learning.

[0047] Fig. 5 is a diagram showing an example of a model use permission setting screen, which is one of the screens for setting a security policy in step 102. The screen shown in Fig. 5 is displayed on the display 25 provided in the client 10. In the screen shown in Fig. 5, "task name" is the name of the task registered by the client 10, and "dataset" is information about the registered data (dataset). Here, the number of data items is included as information about the dataset.

[0048] "Model summary information" is information about the industry and business entered by the user when the client 10 registered the data. "Model usage permission options" specify the range of permitted use of the model, which can be selected and set for each piece of data registered by the client 10. The user can select one of the options to specify the range of federated learning in accordance with their company's security tolerance and business needs. Although options are used here, the present invention is not limited to this, and the user may input the options to specify the range of permitted use.

[0049] In the example shown in FIG. 5, there are five options for "Model Usage Permission Options," and one can be selected using the radio buttons 60. Note that this is just one example, and the number of options is not limited to five, but may be two to four, or six or more. Furthermore, the method for selecting one of the options is not limited to using the radio buttons 60.

[0050] "All Partner Alliance" is an option that allows all alliance partners participating in federated learning to use the model as a federated partner. In other words, all alliance partners are permitted to use the data registered by the client 10. "Cross-industry Partner Alliance" is an option that allows only alliance partners from different industries among the alliance partners participating in federated learning to use the model as a federated partner. By selecting "Cross-industry Partner Alliance," the risk of data being used between organizations in a competitive relationship can be reduced.

[0051] "Internal federation only" is an option that allows only partners of the same company participating in federated learning to use the model in a federated manner. "Individual designation" is an option that allows individual designation of those who are permitted to use the model. This can be used, for example, when only collaborative research partners are permitted to use the model in a federated manner. "Independent use" is an option that prevents partners participating in federated learning from using the model in a federated manner.

[0052] The user selects one of the model use permission options and presses the registration button 61, whereupon the selected range is sent as the permitted use range to the central 11 and registered.

[0053] Specifically, the reception unit 43 receives a selection of whether or not to permit the use of data as a federation partner, i.e., a selection of one of these options by the user. The reception unit 43 transmits information on the selected permitted range of use to the central 11 via the transmission unit 40, and the information is stored in the storage device 12 connected to the central 11 in association with the corresponding data and related information.

[0054] FIG. 6 is a diagram showing an example of a screen for selecting a federation partner in step 104. The screen displays a list of federation partner models (client models) that the client 10 is permitted to use. The list includes information related to the model, such as the industry and business. The list also includes information such as the type of task and the number of data items used when learning the model.

[0055] The list shown in Fig. 6 has an input field 70 in which partners can be selected by checking them with the pointing device 30 or the like. In the example shown in Fig. 6, two partners whose business is "inquiry" have been selected. A partner model selection button 71 for confirming the selection of federation partners is displayed on the screen, and the user can determine the selected partners as partners who can be used in federation by pressing the partner model selection button 71.

[0056] A federation partner can be selected by selecting a model from a list of models that are permitted for use by the client 10. For example, if client A is permitted to use the models of clients B to D, client A can select a federation partner from among clients B to D.

[0057] The reception unit 43 requests a screen for selecting an alliance partner from the central 11 via the transmission unit 40. Based on the selected information, the central 11 searches for models that the client 10 is permitted to use, and transmits a list of the found models together with related information associated with each model to the client 10. The reception unit 43 receives the list of models and related information via the reception unit 41, and displays a list of related information associated with the models as a selectable list of partner models permitted to use. The reception unit 43 then receives the user's selection of partners permitted to use, and transmits information about the selected partners to the central 11 via the transmission unit 40.

[0058] The central 11 stores the information of the selected partner in association with the data registered by the client 10.

[0059] By the way, if there are a sufficient number of models in the list that can be selected as coalition partners, you can choose from that list, but if there are not enough models, you may want to apply to use a model that you are not permitted to use.

[0060] Therefore, the client 10 can display a screen for requesting the use of a model for which use is not permitted, and can request the use of that model from other clients who have registered that model.

[0061] FIG. 7 is a diagram showing an example of a screen for applying to use a model. The screen displays a list of models of federated partners that are registered in the information processing system and that are not permitted for use by the client 10. The list includes information related to the model, such as the industry and business. The list also includes information such as the type of task used when learning the model and the number of data.

[0062] The list shown in Fig. 7 has an input field 80 in which a partner can be selected by checking the box with the pointing device 30 or the like. In the example shown in Fig. 7, two partners whose business is "survey" have been selected. A model usage request button 81 for confirming the usage request to the alliance partner is displayed on the screen, and the user can apply to use the model to the selected partner by pressing the model usage request button 81.

[0063] When the model usage application button 81 is pressed, a screen for approval or denial is sent to the selected partner, and the screen is displayed to the selected partner. The model also stores the IP address, etc., of each partner in association with it. Therefore, the above-mentioned screen for approval or denial can be sent to the selected partner using the IP address, etc., associated with the selected partner's model.

[0064] The reception unit 43 requests the central 11 via the transmission unit 40 a screen for submitting a usage application for a client model that is not permitted for use. Based on the information about the selected partner, the central 11 searches for models that are not permitted for use by the client 10, and transmits a list of the found models together with related information associated with each model to the client 10. The reception unit 43 receives the list of models and related information via the reception unit 41, and displays a list of partner models that are not permitted for use, as well as a list of related information associated with the models, selectable as a list of related information that can be applied for use. The reception unit 43 then receives the user's selection of a partner for whom a usage application is to be made, and transmits information about the selected partner for whom a usage application is to be made to the central 11 via the transmission unit 40.

[0065] The central 11 stores the information of the selected partner who is the target of the usage application in association with the data registered by the client 10.

[0066] Fig. 8 is a diagram showing an example of a screen for approving or denying a request to use a model. The screen displays model information such as related information when setting the security policy for the model (client model), task name, and data set information. The screen also displays information about the client 10 that has made the request to use the model. The information about the client 10 is summary information, such as information about the industry.

[0067] The screen includes an Approve button 90 to approve the usage request and a Deny button 91 to deny it. When the Approve button 90 is pressed, the usage of the requested model is permitted. On the other hand, when the Deny button 91 is pressed, the usage of the requested model remains denied.

[0068] The screen can include other information, such as the intended use of the model, which can be used as a basis for deciding whether to approve or deny the use of the model.

[0069] The use application sent from the client 10 is received by the central 11, which creates an approval request screen for each partner based on the information of the partner selected in the use application and sends it to each partner. Each partner acquires the approval request screen using the receiving unit 41 as an acquisition unit, and can accept the selection of whether to approve the use application using the accepting unit 43. Each partner transmits whether or not to approve to the central 11, and the central 11 changes the scope of each partner's permitted use of the model based on whether or not to approve.

[0070] 9 is a flowchart showing an example of a process for constructing a model through federated learning. The process for constructing a model through federated learning is executed by the information processing system and starts at step 200. In step 201, the central 11 generates an initial model and distributes it to each client 10.

[0071] In step 202, the client 10 receives the initial model distributed from the central 11, learns the initial model using its own data (client data), and updates the model. In step 203, the central 11 receives the model learned by the client 10 (client model). The client 10 can obtain evaluation results of the model when learning the client model, and can transmit the evaluation results together with the model to the central 11. The evaluation results can include, for example, accuracy rate and recall rate. Note that this is just one example, and the evaluation results are not limited to accuracy rate and the like. The client 10 may also transmit the output results of the model for specific data, rather than the model itself.

[0072] In step 204, the central 11 stores the acquired model in the storage device 12 as a distributable model so that it can be applied after model development. In this case, the central 11 may store only the best model with the highest evaluation in the federated learning cycle identified using the received evaluation results. Here, the model is stored in the storage device 12 connected to the central 11, but this is not limited to this, and the model may also be stored in the storage device 14 connected to the client 10.

[0073] In step 205, the central 11 integrates the models acquired from each client 10 and stored in the storage device 14 to update the central model. If a model is stored in the storage device 14 connected to the client 10, the central 11 acquires the model from the client 10 and updates the central model by creating a new central model using that model or the output of that model. At this time, the central 11 can output the evaluation results of the central model.

[0074] In step 206, the central 11 stores the updated central model in the storage device 12. At that time, the central 11 may store only the best model with the highest evaluation in the federated learning cycle identified using the evaluation results of the central model.

[0075] In step 207, the central 11 determines whether the update termination condition for the model is met. The update termination condition may be a predetermined number of times, a convergence condition, or the like. If the update termination condition is not met, the process proceeds to step 208, where the central 11 distributes the updated central model to each client 10. Then, the process returns to step 202, and the processes of steps 202 to 206 are repeated until the termination condition is met.

[0076] If the termination conditions are met in step 207, the process proceeds to step 209, where the client 10 selects the model that is best suited to their task or business from the central models constructed by federated learning and deploys it. Deploying means implementing the model in the environment in which it will be used and making the model available for use. In step 210, the client 10 applies the selected model to the task or business, and in step 211, the process ends.

[0077] As explained above, each client 10 can freely adjust the security level of the model, which enables federated learning to be performed in a security environment that meets the security requirements and business needs of each client. In addition, by selecting a partner related to the content to be learned using the client 10's model, efficient and purposeful federated learning can be achieved.

[0078] In addition, by applying for and being approved to use data that the client 10 is not permitted to use, the client 10 can train the model using more data sources, thereby improving the accuracy and applicability of the model. Furthermore, by making data available through this application and approval, transparency in the federated learning process is improved, making it easier for the client 10 to actively participate in federated learning.

[0079] The information processing device, information processing system, information processing method, and program of the present invention have been described above using the above-mentioned embodiments. However, the present invention is not limited to the above-mentioned embodiments, and other modifications, additions, changes, deletions, and other changes can be made within the scope of what one skilled in the art can conceive. Furthermore, any embodiment is within the scope of the present invention as long as it achieves the functions and effects of the present invention. [Explanation of symbols]

[0080] 10...client, 11...central, 12...storage device, 13...network, 14...storage device, 15...information processing device, 16...network, 17...storage device, 20...CPU, 21...ROM, 22...RAM, 23...HD, 24...HDD controller, 25...display, 26...external device connection I / F, 27...network I / F, 28...bus line, 29...keyboard, 30...pointing device, 31...DVD-RW drive , 32...media I / F, 33...DVD-RW, 34...recording medium, 40...sending unit, 41...receiving unit, 42...request unit, 43...accepting unit, 44...learning unit, 50...sending unit, 51...receiving unit, 52...registration processing unit, 53...updating unit, 54...determination unit, 60...radio button, 61...registration button, 70...input field, 71...partner model selection button, 80...input field, 81...model usage application button, 90...approval button, 91...denial button [Prior art documents] [Patent documents]

[0081] [Patent Document 1] Japanese Patent Publication No. 2022-76277

Claims

1. An information processing device used to integrate local models by federated learning to construct a global model, a registration means for registering data of the local model obtained by machine learning using a dataset held by the information processing device; a receiving means for receiving a designation of a range of permitted use of the registered data; An information processing device comprising:

2. The information processing apparatus according to claim 1 , wherein the accepting unit accepts a selection of whether or not to permit the use of the data by the partner performing the federated learning.

3. an acquisition means for acquiring related information associated with data of a local model that is permitted to be used among data of a local model that has been machine-learned in one or more other information processing devices that are registered in the registration destination; The information processing apparatus according to claim 2 , wherein the accepting unit displays the acquired related information selectably as related information of the partner who is permitted to use the data.

4. The information processing device according to claim 1 or 2, wherein the reception means receives a selection of whether or not to apply for use of local model data that has been machine-learned in one or more other information processing devices registered as the registration destination and for which use is not permitted.

5. an acquisition means for acquiring related information associated with the data of the local model for which use is not permitted; The information processing apparatus according to claim 4 , wherein the accepting unit displays the acquired related information in a selectable manner as related information associated with data of a local model for which a use application can be made.

6. an acquisition means for acquiring a use application for the data from another information processing device; The information processing apparatus according to claim 4 , wherein said accepting means accepts a selection of whether or not to approve said application for use.

7. An information processing system that integrates local models by federated learning to construct a global model, The system includes one or more information processing devices that perform machine learning of the local models using data sets that the information processing devices hold, and a management device that integrates the machine-learned local models to construct a global model, The information processing device, a registration means for registering data of the local model obtained by machine learning; a receiving means for receiving a designation of a range of permitted use of the registered data; An information processing system comprising:

8. The information processing system according to claim 7 , wherein the accepting unit accepts a selection of whether or not to permit the use of the data by the partner performing the federated learning.

9. The information processing system according to claim 7 or 8, wherein the reception means receives a selection of whether or not to apply for use of local model data that is not permitted for use among data of local models that have been machine-learned in one or more other information processing devices registered in the management device.

10. 9. The information processing system according to claim 7, wherein said accepting unit accepts a selection of whether or not to approve a request for use of said data from another information processing device.

11. The management device updating means for updating the global model using data of a local model that each of the information processing devices has machine-learned; a transmitting means for transmitting data of the global model, which satisfies an update termination condition and has been updated, to the information processing device in response to the selection of the global model from the information processing device; 9. The information processing system according to claim 7, further comprising:

12. 1. An information processing method executed by an information processing device for integrating local models by federated learning to construct a global model, comprising: registering data of the local model obtained by machine learning using a dataset held by the information processing device; receiving a designation of a range of permitted use of the registered data; An information processing method, including:

13. The information processing method according to claim 12 , further comprising the step of accepting a selection of whether or not to permit the use of the data as a partner performing the federated learning.

14. The information processing method according to claim 12 or 13, further comprising a step of accepting a selection of whether or not to apply for use of data of a local model that has been machine-learned in one or more other information processing devices registered in the registration destination and that is not permitted for use.

15. 14. The information processing method according to claim 12, further comprising the step of accepting a selection of whether or not to approve a request for use of said data from another information processing device.

16. A program for causing a computer to execute a process executed to integrate local models by federated learning to construct a global model, registering data of the local model obtained by machine learning using a dataset held by the computer; receiving a designation of a range of permitted use of the registered data; A program that executes.

17. 17. The program according to claim 16, which executes a step of accepting a selection of whether or not to permit the use of the data as a partner performing the federated learning.

18. The program according to claim 16 or 17, which executes a step of accepting a selection of whether or not to apply for use of local model data that has been machine-learned in one or more other information processing devices registered in the registration destination and for which use is not permitted.

19. 18. The program according to claim 16, further comprising: receiving a selection of whether to approve a request for use of the data from another information processing apparatus.

Images