Information processing system and computer program

Abstract

An AI processing device (12) associates and stores a training data set and a functional approximator generated based on the training data set. An AI processing device (12) generates a learning history screen that displays a list of a plurality of groups in which a training data set and a functional approximator generated on the basis of the training data set are associated. In the learning history screen, an operation for deriving from existing groups and creating a new group can be input. The AI processing device (12) generates an edit screen which is an edit screen of a new group and which is capable of editing a training data set associated with an existing group.

Description

Technical Field

[0001] The present disclosure relates to data processing technology, and more particularly, to information processing systems and computer programs. Background Art

[0002] In image recognition processing, for example, Deep Learning is used. Deep Learning is known as a machine learning methodology using a multi-layer neural network, such as a convolutional neural network. A convolutional neural network is formed by repeating convolution and pooling of local regions. In addition, a fully convolutional neural network structure has been proposed in which the fully connected layers constituting the convolutional neural network are set as convolutional layers (for example, see Non-Patent Document 1).

[0003] Prior art literature

[0004] Non-patent literature

[0005] Non-patent literature 1: Jonathan Long, Evan Shelhamer, Trevor Darrell, "FullyConvolutional Networks for Semantic Segmentation", The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015, pp.3431-3440 Summary of the Invention

[0006] Technical problem to be solved by the invention

[0007] Training data is closely related to the performance of a function approximator generated through machine learning based on the training data. System developers using function approximators may want to reproduce past training data that produced a high-performance function approximator at a later time.

[0008] The present disclosure is made in view of the above technical problems, and one object of the present disclosure is to provide a technology for efficiently fabricating an auxiliary function approximator.

[0009] Methods used to solve technical problems

[0010] To solve the above-mentioned technical problems, an information processing system according to one aspect of the present disclosure includes: a model generation unit that generates a function approximator through learning based on a training dataset; a storage unit that associates and stores the training dataset and the function approximator generated based on the training dataset; a learning history screen generation unit that generates a learning history screen that displays a list of multiple groups in which the training datasets and the function approximators generated based on the training datasets are associated; and an editing screen generation unit. The learning history screen allows input of an operation to create a new group from an existing group. Upon input of the operation, the editing screen generation unit generates an editing screen for the new group, which allows editing of the training dataset associated with the existing group.

[0011] It should be noted that any combination of the above-described constituent elements and any form of expression of the present disclosure in the form of a device, a method, a computer program, a recording medium storing the computer program, etc. may also be effective as aspects of the present disclosure.

[0012] Effects of the Invention

[0013] According to the technology disclosed in the present invention, efficient fabrication of a function approximator can be assisted. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a diagram showing the configuration of an information processing system according to an embodiment.

[0015] Figure 2 It shows Figure 1 A block diagram of the functional blocks of an AI processing device.

[0016] Figure 3 This is a flowchart showing the operation of the AI ​​processing device related to assisting in the generation of a new AI model.

[0017] Figure 4 It is a diagram showing an example of a learning history screen.

[0018] Figure 5 2 is a diagram showing an example of a new creation window.

[0019] Figure 6 It is a diagram showing an example of a new learning history screen.

[0020] Figure 7 It is a diagram showing an example of an editing screen.

[0021] Figure 8 It is a diagram showing an example of a learning history screen according to a modification. DETAILED DESCRIPTION

[0022] The main body of the device or method disclosed herein includes a computer. The computer realizes the main functions of the device or method disclosed herein by executing a program. The computer has a processor that works according to the program as its main hardware component. The type of processor is not limited as long as it can realize the function by executing the program. The processor is composed of one or more electronic circuits including a semiconductor integrated circuit (IC) or LSI (Large Scale Integration). It is called IC or LSI here, but the name is changed depending on the degree of integration, and it can also be called system LSI, VLSI (Very Large Scale Integration) or USLI (Ultra Large Scale Integration). Field Programmable Gate Array (FPGA) programmed after LSI manufacturing, or reconfigurable logic devices that can be assembled to reconstruct the connection relationship within LSI or the circuit division within LSI can also be used for the same purpose. Multiple electronic circuits can be integrated into one chip or provided in multiple chips. Multiple chips can be integrated into one device or possessed by multiple devices. The program may be stored in a non-volatile recording medium such as a computer-readable ROM (Read Only Memory), an optical disk, or a hard disk drive, or in a volatile storage medium such as a computer-readable RAM (Random Access Memory). The program may be pre-stored in a recording medium or provided to a recording medium or storage medium via a wide area communication network such as the Internet.

[0023] First, an overview of the technology described in the embodiments will be provided. In recent years, function approximators generated through machine learning have been utilized in various fields. For example, image recognition processing based on function approximators is sometimes utilized in the inspection of manufactured products in factories and other places. Function approximators can also be referred to as mathematical models or AI models, and in the embodiments, they are referred to as AI models.

[0024] Training data (in other words, learning data) is closely related to the performance of AI models created through machine learning based on this training data. As AI models are gradually created, their performance deteriorates. System developers using AI models may want to later reproduce past training data that produced high-performing AI models.

[0025] Therefore, the information processing system of the embodiment provides the user with a learning history screen that displays a list of multiple groups in which training data sets are associated with AI models generated based on the training data sets. The training data set of the embodiment is set to include both a training image and an image of a partial area of ​​the training image captured from the training image. The information processing system of the embodiment also provides the user with an editing screen for the new group when an operation is input to derive from an existing group, which is an editing screen for the new group and allows editing of the training data set associated with the existing group. This assists in the efficient creation of AI models.

[0026] The details of the technology of the embodiment will be described. Figure 1 The configuration of an information processing system 10 according to the embodiment is shown. The information processing system 10 includes an AI processing device 12 and a plurality of user terminals 14 . Figure 1 The devices shown are connected via a communication network 16 including a LAN, a WAN, the Internet, and the like.

[0027] The AI ​​processing device 12 is an information processing device that performs processing related to the generation and evaluation of an AI model. The AI ​​processing device 12 of the embodiment is a cloud server that provides data processing services as cloud services.

[0028] Multiple user terminals 14 are information processing devices operated by multiple users who utilize the services of AI processing device 12. For example, these are information processing devices operated by developers of systems using AI models. Multiple user terminals 14 include user terminal 14a, user terminal 14b, and user terminal 14c, each operated by a different user. User terminals 14 may also be PCs, tablets, or smartphones.

[0029] In the embodiment, AI processing device 12 functions as a web server. AI processing device 12 provides web content (HTML data, etc.) related to AI model development or evaluation to user terminal 14. User terminal 14 accesses AI processing device 12 via a web browser. Furthermore, user terminal 14 displays the web content provided by AI processing device 12 (e.g., the AI ​​management screen described later) on a predetermined display.

[0030] Figure 2 It shows Figure 1 This is a block diagram of the functional blocks of the AI ​​processing device 12. Each block shown in the block diagram of this disclosure can be implemented in hardware by components or mechanical devices, such as a computer CPU and memory, and in software by computer programs, etc. However, here, functional blocks implemented through the collaboration of these components are illustrated. Those skilled in the art will understand that these functional blocks can be implemented in various forms through a combination of hardware and software.

[0031] The AI ​​processing device 12 includes a processing unit 20, a storage unit 22, and a communication unit 24. The processing unit 20 performs various data processing related to the development and evaluation of AI models. The storage unit 22 stores data referenced or updated by the processing unit 20. The communication unit 24 communicates with external devices according to a predetermined communication protocol. The processing unit 20 transmits and receives data to and from the user terminal 14 via the communication unit 24.

[0032] The storage unit 22 includes a dataset storage unit 26 and an evaluation data storage unit 28. The dataset storage unit 26 stores various data related to the AI ​​model. The dataset storage unit 26 of the embodiment associates and stores a training dataset (e.g., multiple training images) with an AI model generated based on the training dataset.

[0033] The data group associated with an AI model stored in the data set storage unit 26 is referred to as a "model data set" below. A model data set includes (1) ID, (2) name, (3) training data set, (4) evaluation data set (e.g., multiple evaluation images), (5) information on the type of classification target (also called "judgment category"), (6) model data of the AI ​​model (e.g., a file storing the AI ​​model), (7) AI model status, (8) copy source, and (9) update time. For (8) copy source, the ID of another model data set that is a derivative of the model data set is set.

[0034] The evaluation data storage unit 28 stores data related to the evaluation of the AI ​​model. Specifically, the evaluation data storage unit 28 stores information related to the correctness of the classification of the evaluation data set based on the AI ​​model and evaluation index values ​​related to the classification accuracy of the AI ​​model.

[0035] The processing unit 20 includes an image registration unit 30, a type registration unit 32, a data set setting unit 34, a model generation unit 36, an evaluation unit 38, a management screen generation unit 40, a management screen providing unit 46, and a copying unit 48. The management screen generation unit 40 includes a learning history screen generation unit 42 and an editing screen generation unit 44.

[0036] The functions of the multiple functional blocks of the processing unit 20 may also be implemented in a computer program (referred to herein as an "AI management program"). The AI ​​management program may also be stored on a non-volatile recording medium and installed into the storage device of the AI ​​processing device 12 via the recording medium. Alternatively, the AI ​​management program may be downloaded via a network and installed into the storage device of the AI ​​processing device 12. The processor (CPU, etc.) of the AI ​​processing device 12 may also read and execute the AI ​​management program from the main memory to perform the functions of the multiple functional blocks of the processing unit 20.

[0037] The operation of the information processing system 10 based on the above configuration will be described. In the following description, the web content provided as a user interface by the AI ​​processing device 12 to the user terminal 14 is collectively referred to as the "AI management screen." The AI ​​management screen includes the learning history screen and the editing screen described below.

[0038] User terminal 14 accesses AI processing device 12 via a web browser based on user operations. Management screen generator 40 of AI processing device 12 generates various AI management screens. Management screen provider 46 of AI processing device 12 transmits AI management screen data to user terminal 14. User terminal 14 displays the AI ​​management screen on a designated display. User terminal 14 transmits information or data corresponding to user operations input into the AI ​​management screen to AI processing device 12.

[0039] The image registration unit 30 of the AI ​​processing device 12 stores the plurality of training images and the plurality of evaluation images uploaded from the user terminal 14 as a model dataset identified by a specific ID (hereinafter referred to as a "target model dataset") in the dataset storage unit 26. The type registration unit 32 of the AI ​​processing device 12 stores information on the type of separation target based on the AI ​​model uploaded from the user terminal 14 in the dataset storage unit 26 as data of the target model dataset.

[0040] The management screen providing unit 46 of the AI ​​processing device 12 sends a data set setting screen (not shown) to the user terminal 14 for extracting a data set (hereinafter also referred to as a "training data set") from a registered training image. A training data set can also be referred to as a plurality of images obtained by extracting a portion of a training image. The user terminal 14 displays the data set setting screen on a predetermined display. On the data set setting screen, the user extracts a training data set from a training image and associates the extracted training data set with any of the multiple registered types.

[0041] The user terminal 14 transmits the training dataset corresponding to any of the multiple categories set on the dataset setting screen to the AI ​​processing device 12. The dataset setting unit 34 of the AI ​​processing device 12 receives the training dataset transmitted from the user terminal 14 and stores the training dataset as object model dataset data in the dataset storage unit 26. In practice, the training dataset corresponding to each of the multiple categories that have been registered is uploaded and registered from the user terminal 14 to the AI ​​processing device 12.

[0042] The user terminal 14 transmits data requesting AI model generation to the AI ​​processing device 12 in response to user operations on the AI ​​management screen. The data requesting AI model generation includes designations of a training dataset used in generating the AI ​​model and an evaluation dataset used in evaluating the AI ​​model.

[0043] The model generation unit 36 ​​of the AI ​​processing device 12 generates an AI model, which serves as a function approximator, through machine learning based on the training dataset specified in the request from the user terminal 14. The AI ​​model can be a fully convolutional neural network or a mathematical model different from a neural network (e.g., a decision tree). The model generation unit 36 ​​associates the generated AI model data with the training dataset used to generate the AI ​​model and stores the data in the dataset storage unit 26.

[0044] The evaluation unit 38 of the AI ​​processing device 12 evaluates the AI ​​model generated by the model generation unit 36 ​​based on pre-registered evaluation data (e.g., evaluation images). The AI ​​model can be evaluated using known methods. For example, the evaluation unit 38 can measure the performance of the AI ​​model based on the accuracy of the results of the AI ​​model's classification of the evaluation data. The evaluation unit 38 associates the evaluation results of the AI ​​model with the identification information of the AI ​​model (e.g., the model dataset ID) and stores them in the evaluation data storage unit 28.

[0045] Next, the operations related to assisting in the creation of a new AI model will be described.

[0046] Figure 3 It is a flowchart showing the actions of the AI ​​processing device 12 related to the assistance in the creation of a new AI model. The user terminal 14 sends data requesting a learning history screen to the AI ​​processing device 12 according to the user's operation. After receiving the data requesting a learning history screen (Y of S10), the learning history screen generation unit 42 of the AI ​​processing device 12 generates data for a learning history screen with reference to the model data set stored in the data set storage unit 26 (S11). The management screen providing unit 46 of the AI ​​processing device 12 sends the data for the learning history screen to the user terminal 14 (S12). The user terminal 14 causes a specified display to display the learning history screen. If the AI ​​processing device 12 does not receive data requesting a learning history screen (N of S10), the processing of S11 and S12 is skipped.

[0047] Figure 4 An example of a learning history screen is shown. The learning history screen 50 lists multiple sets (i.e., multiple model data sets) in which training data sets are associated with AI models generated based on the training data sets. One record in the learning history screen 50 corresponds to one model data set.

[0048] The "Status" on the learning history screen 50 corresponds to the (7) AI model status of the model dataset, and is set to, for example, any of: Not Learned (AI model not generated), Learning Completed (AI model generation completed), or Evaluation Completed. The "Name" on the learning history screen 50 corresponds to the (2) Name of the model dataset. The "Copy Source" on the learning history screen 50 corresponds to the (8) Copy Source of the model dataset. The "Update Time" on the learning history screen 50 corresponds to the (9) Update Time of the model dataset.

[0049] The copy source information of the learning history screen 50 is information showing the derivation relationship between a plurality of model data sets. Figure 4 The value of the copy source of the model dataset "20221226-152201" is created by deriving the model dataset from another model dataset "20221226-145954".

[0050] Furthermore, the learning history screen 50 includes an edit button 52, a delete button 54, and an add button 56 as objects that receive user operations. The edit button 52 is selected when editing a model dataset. The delete button 54 is selected when deleting a model dataset. The add button 56 is selected when adding a new model dataset.

[0051] When the user creates a new AI model, he or she selects the add button 56 on the learning history screen 50. The learning history screen 50 is configured so that when the add button 56 is selected, a new creation window 60 is displayed as a pop-up window on the learning history screen 50.

[0052] Figure 5 An example of a new creation window is shown. The new creation window 60 includes an area for inputting a name for a new model dataset. Furthermore, the new creation window 60 includes an object for selecting whether to create a new model dataset without using an existing model dataset or to derive (copy) the new model dataset from an existing model dataset. Furthermore, the new creation window 60 includes an object for selecting a model dataset from which to derive.

[0053] exist Figure 5 In the new creation window 60, the user selects to derive (copy) the existing model dataset to create a new model dataset "20221226-152525". In addition, in the new creation window 60, the user selects the model dataset "20221226-152201" as the derivation source. Then, the user selects the register button in the creation window 60. This series of operations is an operation to derive a new model dataset from an existing model dataset. In response to this user operation, the user terminal 14 sends a new creation request including the name of the new model dataset and information about the derivation source to the AI ​​processing device 12.

[0054] Upon receiving a new creation request from user terminal 14 (Y in S13), copy unit 48 of AI processing device 12 copies the derived model dataset. Copy unit 48 changes the name of the copied model dataset to the name notified by user terminal 14 and stores it in dataset storage unit 26 (S14).

[0055] Specifically, the copying unit 48 stores a new model dataset containing at least (3) a training dataset, (4) an evaluation dataset, (5) information on the type of classification target, and (6) data of the AI ​​model in the dataset storage unit 26. Furthermore, the copying unit 48 associates the model dataset with the model dataset of the derived source. In other words, it records information indicating that the model dataset is derived from the model dataset of the derived source. Specifically, the copying unit 48 sets the ID of the model dataset of the derived source as the (8) copy source of the new model dataset.

[0056] The learning history screen generator 42 of the AI ​​processing device 12 generates data for a new learning history screen 50 that includes a record of the new model dataset (S15). The management screen provider 46 transmits the data for the new learning history screen 50 to the user terminal 14. The user terminal 14 displays the new learning history screen 50 on a predetermined display. If the AI ​​processing device 12 does not receive a new creation request (N in S13), it skips steps S14 and S15.

[0057] Figure 6 An example of a new learning history screen is shown. Figure 6 In the new learning history screen 50, a record of a new model data set whose status is set to "unlearned" is added. In addition, for its copy source, set Figure 5 In the new creation window 60, the model dataset "20221226-152201" is specified as the derived source. To edit the new model dataset, the user selects the edit button 52 on the learning history screen 50. In response to the user's operation, the user terminal 14 transmits data requesting the edit screen for the new model dataset, specifying the ID of the new model dataset.

[0058] When an edit screen is requested from the user terminal 14 (Y in S16), the edit screen generation unit 44 of the AI ​​processing device 12 obtains the contents of the new model dataset (e.g., a training dataset or an evaluation dataset) from the dataset storage unit 26 and generates an edit screen representing the contents of the new model dataset (S17). The edit screen is a user interface that allows the user to set the contents of the new model dataset by editing the contents of the existing model dataset (e.g., a training dataset or an evaluation dataset) that is a derived source. The management screen provision unit 46 sends the edit screen data to the user terminal 14 (S18). The user terminal 14 displays the edit screen on a predetermined display. If no edit screen request is received from the user terminal 14 (N in S16), the processing of S17 and S18 is skipped.

[0059] As a variation, the editing screen generator 44 may be included in the user terminal 14. In this variation, the user terminal 14 may transmit data requesting new model dataset data, specifying the ID of the new model dataset, to the AI ​​processing device 12 in response to a user operation. The AI ​​processing device 12 may also transmit the contents of the new model dataset stored in the dataset storage unit 26 to the user terminal 14. The editing screen generator of the user terminal 14 may also generate and display an editing screen provided by the AI ​​processing device 12 that shows the contents of the new model dataset.

[0060] Figure 7 An example of an editing screen is shown. Editing screen 70 in the figure is related to editing a new model dataset derived from an existing model dataset, and shows the editing screen for a training dataset 72. In editing screen 70, a training dataset 72 associated with an existing model dataset can be edited. For example, by selecting Add button 74, a user can add new training data not already in the existing model dataset to the model dataset being edited. Furthermore, by selecting Delete button 76, a user can delete at least a portion of the training data registered in the existing model dataset from the model dataset being edited.

[0061] The user terminal 14 transmits data indicating the editing results (e.g., addition of a new training dataset, deletion of a training dataset integrated from a derived source) displayed on the editing screen 70 to the AI ​​processing device 12. The image registration unit 30 and dataset setting unit 34 of the AI ​​processing device 12 modify the data of the new model dataset stored in the dataset storage unit 26 to reflect the editing results displayed on the editing screen 70.

[0062] In response to a request from the user terminal 14, the model generation unit 36 ​​of the AI ​​processing device 12 generates a new AI model through machine learning based on the new model dataset stored in the dataset storage unit 26. The model generation unit 36 ​​stores the data for the new AI model in a new model dataset in the dataset storage unit 26. Specifically, the dataset storage unit 26 associates and stores the edited training dataset with the AI ​​model generated based on the training dataset as the data for the new model dataset. Furthermore, as previously mentioned, the dataset storage unit 26 stores information indicating that the new model dataset is derived from an existing model dataset.

[0063] The evaluation unit 38 of the AI ​​processing device 12 evaluates the new AI model, associates the evaluation results with the new model dataset, and stores them in the evaluation data storage unit 28. After the user confirms the evaluation results of the new AI model, they can create a new model dataset derived from the new model dataset, thereby generating an even newer AI model. Furthermore, the user can confirm the evaluation results of the new AI model, return to the derived model dataset, and create a model dataset for another system.

[0064] The information processing system 10 according to the embodiment makes it easy to reuse training datasets used in the creation of previous AI models or evaluation datasets, thereby assisting in the efficient creation or improvement of AI models. Furthermore, the information processing system 10 can assist in the efficient creation of new AI models by providing users with the derivational relationships between multiple model datasets. For example, this system can assist users in selecting appropriate derivation sources when creating new AI models.

[0065] While the present disclosure has been described above based on the embodiments, those skilled in the art will appreciate that the embodiments are merely illustrative and that various modifications can be made to the combinations of the components or processes of the embodiments, and such modifications are also within the scope of the present disclosure.

[0066] Modifications will be described. Figure 8 An example of a learning history screen according to a modification is shown. Figure 8 The learning history screen 50 includes an image that vividly displays the derivation relationship between the model datasets stored in the dataset storage unit 26 of the AI ​​processing device 12. The learning history screen generation unit 42 of the AI ​​processing device 12 generates the learning history screen 50 based on the derivation relationship between the multiple model datasets stored in the dataset storage unit 26. The learning history screen 50 connects the objects in the multiple model datasets with arrows, with the starting point of the arrow being the derivation source and the end point of the arrow being the derivation target. Figure 8 The learning history screen 50 shows that the parent model derives child model 1 and child model 2, and the grandchild model derives from the child model 1. According to this modification, a learning history screen can be provided that allows the user to intuitively grasp the derivation relationship between multiple model data sets.

[0067] Other variations are described below. The AI ​​processing device 12 in the above embodiment is configured as a cloud server, but as a variation, the AI ​​processing device 12 may also be a local server. Furthermore, the functions of the AI ​​processing device 12 in the embodiment may be distributed and installed in multiple information processing devices. In this case, the multiple information processing devices communicate with each other and, by collaborating as a system, perform the same processing as the AI ​​processing device 12 in the embodiment. Furthermore, at least a portion of the functions of the AI ​​processing device 12 in the embodiment may also be installed in an application that operates on the user terminal 14.

[0068] Any combination of the above-mentioned embodiments and modifications is still useful as an embodiment of the present disclosure. The new embodiment generated by the combination has the effects of the combined embodiments and modifications. In addition, it should be understood by those skilled in the art that the functions to be achieved by each component described in the claims can be achieved by the individual components shown in the embodiments and modifications or by their collaboration.

[0069] <Note>

[0070] Based on the description of the above-mentioned embodiment and modified examples, the following technology is disclosed.

[0071] [Technology 1]

[0072] An information processing system, comprising:

[0073] The model generation unit generates a function approximator by learning based on the training data set.

[0074] a storage unit that associates and stores the training data set with a function approximator generated based on the training data set,

[0075] a learning history screen generating unit for generating a learning history screen that displays a list of a plurality of sets in which the training data set and the function approximators generated based on the training data set are associated; and

[0076] Editing screen generation unit;

[0077] In the learning history screen, an operation to create a new group by deriving from an existing group can be input.

[0078] When the operation is input, the editing screen generating unit generates an editing screen for the new group, which is an editing screen capable of editing the training data set associated with the existing group.

[0079] According to this information processing system, it is easy to reuse a training data set used in the creation of a function approximator in the past, and it is possible to assist in the efficient creation of a function approximator.

[0080] [Technique 2]

[0081] According to the information processing system described in technology 1,

[0082] The editing screen is configured to enable addition of new training data to the training data set associated with the existing group.

[0083] This information processing system can assist in editing (for example, adding new training data) a training data set used in creating a previous function approximator, and can assist in efficiently creating a new function approximator.

[0084] [Technique 3]

[0085] The information processing system according to technique 1 or 2,

[0086] The storage unit stores, as data of the new group, an edited training data set and a function approximator generated based on the edited training data set, and stores information indicating that the new group is derived from the existing group.

[0087] According to this information processing system, it is possible to store information indicating the derivational relationship between a training data set and a plurality of groups including a function approximator.

[0088] [Technique 4]

[0089] The information processing system according to any one of techniques 1 to 3,

[0090] The learning history screen generating unit generates a learning history screen including information indicating a derivation relationship between the plurality of groups.

[0091] This information processing system assists users in efficiently creating new function approximators by providing them with information about derivation relationships between multiple groups of training datasets and function approximators. For example, it assists users in selecting appropriate derivation sources when creating new function approximators.

[0092] [Technique 5]

[0093] A computer program that causes a computer to:

[0094] Generate a function approximator by learning based on the training data set,

[0095] Associating the training data set with a function approximator generated based on the training data set and storing the function approximator in a storage unit,

[0096] generating a learning history screen that displays a list of a plurality of sets associated with the training data set and the function approximators generated based on the training data set;

[0097] In the learning history screen, an operation of creating a new group by deriving from an existing group can be input.

[0098] When the operation is input, the computer is caused to further execute: generating an editing screen, which is an editing screen of the new group and capable of editing the training data set associated with the existing group.

[0099] This computer program facilitates the reuse of a training data set used in the creation of a function approximator in the past, and enables a computer to implement processing that assists in the efficient creation of a function approximator.

[0100] Industrial Availability

[0101] The technology of the present disclosure can be applied to an information processing system or an information processing device.

[0102] Description of Reference Numerals

[0103] 10 Information processing system, 12 AI processing device, 14 User terminal, 26 Data set storage unit, 36 Model generation unit, 42 Learning history screen generation unit, 44 Editing screen generation unit.

Claims

1. An information processing system comprising: The model generation unit generates a function approximator by learning based on the training data set. a storage unit that associates and stores the training data set with a function approximator generated based on the training data set, a learning history screen generating unit for generating a learning history screen that displays a list of a plurality of sets in which the training data set and the function approximators generated based on the training data set are associated; and Editing screen generation unit; In the learning history screen, an operation to create a new group by deriving from an existing group can be input. The editing screen generating unit generates an editing screen for the new group, which is capable of editing the training data set associated with the existing group.

2. The information processing system according to claim 1, The editing screen is configured to enable addition of new training data to the training data set associated with the existing group.

3. The information processing system according to claim 1 or 2, The storage unit stores, as data of the new group, an edited training data set and a function approximator generated based on the edited training data set, and stores information indicating that the new group is derived from the existing group.

4. The information processing system according to claim 1 or 2, The learning history screen generating unit generates a learning history screen including information indicating a derivation relationship between the plurality of groups.

5. A computer program that causes a computer to: Generate a function approximator by learning based on the training data set, Associating the training data set with a function approximator generated based on the training data set and storing the function approximator in a storage unit, generating a learning history screen that displays a list of a plurality of sets associated with the training data set and the function approximators generated based on the training data set; In the learning history screen, an operation of creating a new group by deriving from an existing group can be input. When the operation is input, the computer is caused to further execute: generating an editing screen, which is an editing screen of the new group and capable of editing the training data set associated with the existing group.

Images