WO2022201256A1

WO2022201256A1 - Information processing device, information processing system, information processing method, and program

Info

Publication number: WO2022201256A1
Application number: PCT/JP2021/011772
Authority: WO
Inventors: 智之吉山
Original assignee: 日本電気株式会社
Priority date: 2021-03-22
Filing date: 2021-03-22
Publication date: 2022-09-29
Also published as: JPWO2022201256A1

Abstract

The present invention generates a data generating model capable of generating more diversified data even in a condition where it is difficult to acquire training data retained by another information processing device. An information processing device (10) is provided with: a training data acquisition unit (101) that acquires training data; a generated model information acquisition unit (102) for acquiring first generated model information which is relating to a first data generating model and in which second generated model information relating to a second data generating model different from the first data generating model is reflected; and a data generating model training unit (103) that trains, by using the training data, the first data generating model defined with reference to the first generated model information.

Description

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

The present invention relates to an information processing device, an information processing system, an information processing method, and a program.

A technology is known that uses machine learning to generate a data generation model that generates data. For example, in Non-Patent Document 1, a data generation model that generates data is trained against both a prediction model that predicts the class of data and a discriminant model that discriminates the truth of data. A technique for generating is described. Data generated by such a data generation model is used, for example, as learning data for an inference model.

However, with the technology described in Non-Patent Document 1, there is room for improvement in the diversity of data generated by the data generation model. For example, if learning data can be acquired from another information processing apparatus when there is a shortage of learning data, it may be possible to improve the learning efficiency of the data generation model. However, even in such a case, in practice, it may be difficult to acquire learning data held by other information processing apparatuses, for example, in consideration of privacy.

One aspect of the present invention has been made in view of the above problem, and an example of its purpose is to improve diversity even in situations where it is difficult to acquire learning data held by other information processing apparatuses. To provide a technique for generating a data generation model capable of generating certain data.

An information processing apparatus according to one aspect of the present invention includes learning data acquisition means for acquiring learning data; and first generative model information relating to a first data generative model, a generative model information acquiring means for acquiring first generative model information reflecting second generative model information relating to a different second data generative model; Data generation model learning means for learning the first data generation model using the learning data.

An information processing apparatus according to one aspect of the present invention includes seed data acquisition means for acquiring seed data and data generation means for generating data using the seed data, wherein the data generation means generates first data First generative model information about a generative model that reflects second generative model information about a second data generative model that is different from the first data generative model with reference to the first generative model information Data generation is performed using the defined first data generation model.

An information processing apparatus according to one aspect of the present invention includes first generative model information about a first data generative model and second generative model information about a second data generative model different from the first data generative model. and generative model information updating means for generating updated generative model information by referring to the first generative model information and the second generative model information. there is

An information processing system according to one aspect of the present invention is an information processing system including a first information processing device and a server, wherein the first information processing device includes learning data acquisition means for acquiring learning data. and first generation model information relating to the first data generation model, which reflects second generation model information relating to a second data generation model different from the first data generation model a generative model information acquiring means for acquiring model information from the server; and a learning data acquiring means for acquiring the first data generating model defined by referring to the first generative model information. data generation model learning means for learning using data; and transmission means for transmitting third generation model information related to the first data generation model learned by the data generation model learning means; is a generative model information acquiring means for acquiring the third generative model information and fourth generative model information related to the second data generative model after learning; the third generative model information; Generative model information updating means for generating updated generative model information by referring to the generative model information of 4; and transmitting means for transmitting the updated generative model information to the first information processing device. ing.

An information processing method according to one aspect of the present invention includes obtaining learning data, obtaining first generative model information related to a first data generative model, wherein the second data generative model differs from the first data generative model. Acquiring first generative model information reflecting second generative model information relating to a data generation model, and obtaining the first data generation model defined by referring to the first generative model information, It includes learning using the learning data.

An information processing method according to one aspect of the present invention includes obtaining seed data and generating data using the seed data, wherein the step of generating data includes: First generative model information, defined by referring to first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model Data generation is performed using the first data generation model.

An information processing method according to one aspect of the present invention includes first generative model information about a first data generative model, and second generative model information about a second data generative model different from the first data generative model. and referring to the first generative model information and the second generative model information to generate updated generative model information.

A program according to one aspect of the present invention is a program for causing a computer to function as an information processing apparatus, the computer comprising: learning data acquisition means for acquiring learning data; Generative model information acquisition means for acquiring first generative model information, which is one piece of generative model information and reflects second generative model information related to a second data generation model different from the first data generation model. and data generation model learning means for learning the first data generation model defined by referring to the first generation model information using the learning data.

A program according to one aspect of the present invention is a program for causing a computer to function as an information processing device, the computer comprising seed data acquisition means for acquiring seed data and generating data using the seed data. The data generating means is first generative model information relating to a first data generating model, and second generative model information relating to a second data generating model different from the first data generating model is reflected. It functions as data generation means for generating data using a first data generation model defined by referring to the first generation model information.

A program according to one aspect of the present invention is a program for causing a computer to function as an information processing apparatus, the computer being configured to generate first generative model information relating to a first data generative model and the first data generative model. a generative model information acquiring means for acquiring second generative model information related to a second data generative model different from the model; and generative model information updating means for generating the generative model information.

According to one aspect of the present invention, it is possible to efficiently learn a data generation model even in a situation where it is difficult to acquire learning data possessed by another information processing apparatus.

1 is a block diagram showing the configuration of an information processing apparatus according to Exemplary Embodiment 1; FIG. FIG. 3 is a flow diagram showing the flow of an information processing method according to exemplary embodiment 1; FIG. 9 is a block diagram showing the configuration of an information processing apparatus according to Exemplary Embodiment 2; FIG. 10 is a flow diagram showing the flow of an information processing method according to exemplary embodiment 2; FIG. 12 is a block diagram showing the configuration of an information processing apparatus according to exemplary embodiment 3; FIG. 11 is a flow diagram showing the flow of an information processing method according to exemplary embodiment 3; FIG. 12 is a block diagram showing the configuration of an information processing system according to Exemplary Embodiment 4; FIG. 12 is a flow diagram showing the flow of an information processing method according to exemplary embodiment 4; FIG. 12 is a block diagram showing the configuration of an information processing system according to Exemplary Embodiment 5; FIG. 12 is a flow diagram illustrating the flow of an information processing method according to exemplary embodiment 5; FIG. 12 is a flow diagram illustrating the flow of another information processing method according to exemplary embodiment 5; FIG. 13 is a schematic diagram illustrating a specific example of a data generation model in exemplary embodiment 5; FIG. 21 is a schematic diagram illustrating another specific example of the data generation model in exemplary embodiment 5; FIG. 21 is a schematic diagram illustrating still another specific example of the data generation model in exemplary embodiment 5; FIG. 12 is a block diagram showing the configuration of an information processing system according to Exemplary Embodiment 6; FIG. 11 is a flow diagram illustrating the flow of an information processing method according to exemplary embodiment 6; FIG. 12 is a block diagram showing the configuration of an information processing system according to exemplary embodiment 7; FIG. 22 is a block diagram showing the configuration of an information processing system according to exemplary embodiment 8; FIG. 22 is a block diagram showing the configuration of an information processing system according to Exemplary Embodiment 9; FIG. 22 is a diagram showing the configuration of a learning data acquisition unit in exemplary embodiment 10; 1 is a block diagram showing an example of hardware configuration of an information processing device and a server in each exemplary embodiment of the present invention; FIG.

[Exemplary embodiment 1]
A first exemplary embodiment of the invention will now be described in detail with reference to the drawings. This exemplary embodiment is the basis for the exemplary embodiments described later.

The information processing apparatus 10 according to this exemplary embodiment uses learning data (hereinafter also referred to as generative model learning data) to learn a data generation model, and can also be referred to as a data generation model learning apparatus. It is. Here, how the data generated by the data generation model according to this exemplary embodiment is used after generation does not limit this exemplary embodiment, but as an example, the data is It can be used to train an inference model.

For example, due to privacy concerns, etc., a situation may arise in which it is difficult to obtain sufficient inference model training data for training an inference model. In such a situation, the information processing device 10 according to this exemplary embodiment can be used as a device that generates inference model learning data for learning the inference model.

On the other hand, even when acquiring generative model learning data, there may be situations where it is difficult to acquire sufficient generative model learning data due to privacy concerns. For example, consider a situation in which there are a plurality of information processing apparatuses having the same configuration as the information processing apparatus 10 . If the data generation model can be learned by directly using the generative model learning data acquired by the plurality of information processing devices, the learning efficiency of the data generation model can be improved. However, in practice, there may arise a situation in which the information processing apparatus 10 should avoid acquiring the generative model learning data acquired by another information processing apparatus in consideration of privacy.

The information processing apparatus 10 according to this exemplary embodiment is configured so that the data generation model can be efficiently learned while keeping consideration for privacy even in such a situation.

The information processing device 10 according to this exemplary embodiment will be described in more detail below.

<Configuration of information processing device>
A configuration of an information processing apparatus 10 according to this exemplary embodiment will be described with reference to FIG. FIG. 1 is a block diagram showing the configuration of an information processing device 10. As shown in FIG. As shown in FIG. 1 , the information processing apparatus 10 includes a learning data acquisition unit 101 , a generative model information acquisition unit 102 , and a data generation model learning unit 103 .

Here, the learning data acquisition unit 101 is an example of the learning data acquisition means according to this exemplary embodiment. Also, the generative model information acquisition unit 102 is an example of a generative model information acquisition unit according to this exemplary embodiment. Also, the data generation model learning unit 103 is an example of the data generation model learning means according to this exemplary embodiment.

(learning data acquisition unit)
The learning data acquisition unit 101 acquires learning data (generative model learning data). The generative model learning data is data for learning a first data generative model, which will be described later.

(generative model information acquisition unit)
The generative model information acquisition unit 102 acquires first generative model information regarding a first data generative model, which will be described later.

(First data generation model)
A first data generation model is a model that generates data. For example, the first data generation model inputs seed data and outputs data. The data generated by the first data generation model is used, for example, as the inference model learning data described above. However, the use of data generated by the first data generation model is not limited to this.

(First generative model information)
The first generative model information includes, for example, various parameters that define the first data generative model. In this exemplary embodiment, the first generative model information reflects second generative model information relating to a second data generative model different from the first data generative model. This makes it possible to indirectly refer to other generative model learning data without directly acquiring generative model learning data possessed by another information processing apparatus, which will be described later. As an example, when the data generation model learning unit 103, which will be described later, is configured to sequentially learn the first data generation model, the first generation model acquired by the generative model information acquisition unit 102 at a certain point in time The information includes first generative model information that defined the first generative model information in the past before the certain point in time, and second generative model information that defined the second generative model information in the past than the certain point in time. It includes information obtained by referring to the generative model information of

(Second data generation model)
The second data generation model is a model different from the first data generation model, and is a model that generates data according to input data. For example, the second generative model is learned using other generative model learning data in an information processing apparatus different from the information processing apparatus 10 that generates the first generative model. The other information processing device may be, for example, a device configured similarly to the information processing device 10 . Also, the second data generation model may be configured to be learned sequentially.

(Second generative model information)
The second generative model information includes, for example, various parameters that define the first data generative model.

Note that the above-described first generative model information may reflect other generative model information relating to another data generation model that is different from either the first data generation model or the second data generation model. Further, it may be configured to reflect other generative model information.

(Data generation model learning part)
The data generation model learning unit 103 learns the first data generation model defined by referring to the first model information acquired by the generative model information acquisition unit 102, using the generative model learning data as teacher data. .

Here, as the first generative model, as an example, a generative model constructed using a neural network, such as an autoencoder or a generative adversarial network, can be used, but this limits this exemplary embodiment. Generative models with other configurations can also be used.

The first model information that defines the first data generation model after learning by the data generation model learning unit 103 is, for example, a second data generation model that defines a second data generation model that is different from the first data generation model. is used to generate updated model information by being combined with the model information of .

<Flow of information processing method>
An information processing method S10 executed by the information processing apparatus 10 configured as described above will be described with reference to FIG. FIG. 2 is a flow diagram showing the flow of the information processing method S10. As shown in FIG. 2, the information processing method S10 includes steps S101 to S103.

(Step S101)
In step S101, the learning data acquisition unit 101 acquires generative model learning data.

(Step S102)
In step S102, the generative model information acquisition unit 102 acquires first generative model information.

(Step S103)
In step S103, the data generation model learning unit 103 creates a first data generation model defined by referring to the first model information acquired by the generative model information acquisition unit 102, using the data for generative model learning as teacher data. learn by using

The information processing device 10 may sequentially learn the first generative model by sequentially executing the information processing method S10 described above. As an example, the first generative model information acquired by the generative model information acquiring unit 102 in step S102 of the information processing method S10 executed for the n-th time (n is a natural number) includes (i) step S103 before the n-1th time and (ii) the second generative model that defined the second generative model information before the n-1 time. Information obtained by taking a weighted average with information is included.

<Effects of this exemplary embodiment>
As described above, in the information processing device 10 according to this exemplary embodiment,
・ Acquire learning data (generative model learning data) of the first data generation model,
A first generative model that is first generative model information about a first data generative model and that reflects second generative model information about a second data generative model that is different from the first data generative model get information,
・Learning a first data generation model defined by referring to the acquired first generative model information using learning data (generative model learning data);
configuration is adopted.

According to the above configuration, without directly acquiring the other generative model learning data used by the other information processing device that generated the second generative model, the other generative model learning data is indirectly acquired. can be used to generate a first data generation model. In other words, data with diversity is generated in the same way as when training is performed by directly using both the data for generative model learning acquired by the own device and the data for generative model learning acquired by another information processing device. A possible first data generation model can be generated.

As a result, even in a situation where the information processing device 10 should avoid acquiring other generative model learning data used in other information processing devices, the first can generate a data generation model of In addition, as an example, when the data generated by such a first data generation model is used for learning an inference model, the diversity of the inference model learning data can be increased, and more accurate and diverse data can be obtained. It is possible to obtain an inference model that can correspond to

The learning performed by the information processing apparatus 10 according to this exemplary embodiment is also called associative learning as an example, but this exemplary embodiment is not limited by this wording.

[Exemplary embodiment 2]
A second exemplary embodiment of the invention will now be described in detail with reference to the drawings. The information processing device 20 according to this exemplary embodiment is a device that generates data using the first data generation model generated by the information processing device 10 according to the first exemplary embodiment. As an example, the data generated by the information processing device 20 may be data to be used as teacher data in learning an inference model. However, the use of the data generated by the information processing device 20 is not limited to this.

The information processing device 20 according to this exemplary embodiment will be described in more detail below.

<Configuration of information processing device>
The configuration of the information processing device 20 according to this exemplary embodiment will be described with reference to FIG. FIG. 3 is a block diagram showing the configuration of the information processing device 20. As shown in FIG. As shown in FIG. 3 , the information processing device 20 includes a seed data acquisition section 201 and a data generation section 202 .

Here, the seed data acquisition unit 201 is an example of seed data acquisition means according to this exemplary embodiment. Also, the data generator 202 is an example of a data generator according to this exemplary embodiment.

(Seed data acquisition part)
The seed data acquisition unit 201 acquires seed data. Seed data is data to be input to the first data generation model. For example, the seed data acquisition unit 201 may acquire by generating seed data, or generated by another functional block (not shown) included in the device or another device (not shown) Seed data may be obtained. A specific example of seed data is random noise data, which is not a limitation of this exemplary embodiment. Also, the name seed data does not limit this exemplary embodiment in any way.

(data generator)
The data generation unit 202 generates data using seed data and the first data generation model. Specifically, the data generation unit 202 generates data by inputting seed data into the first data generation model. The first data generation model is defined with reference to the first generation model information. The first generative model information reflects second generative model information relating to a second data generative model different from the first data generative model. Details of the first data generation model, the first generative model information, the second data generation model, and the second generative model information are as described in the first exemplary embodiment.

<Flow of information processing method>
The information processing method S20 executed by the information processing apparatus 20 configured as described above will be described with reference to FIG. FIG. 4 is a flow diagram showing the flow of the information processing method S20. As shown in FIG. 4, the information processing method S20 includes steps S201 to S203.

(Step S201)
In step S201, the seed data acquisition unit 201 acquires seed data. For example, the seed data acquisition unit 201 may acquire randomly generated data as seed data.

(Step S202)
In step S202, the data generation unit 202 generates data by inputting seed data into the first data generation model. For example, the memory (not shown) of the information processing apparatus 20 stores first generative model information regarding the first data generative model generated by the information processing apparatus 10 according to the first exemplary embodiment. The data generation unit 202 generates data using a first data generation model defined by referring to the first generation model information.

<Effects of this exemplary embodiment>
As described above, in the information processing device 20 according to this exemplary embodiment,
・Obtain seed data,
- Generate data using the acquired seed data and the first data generation model.
- the first data generative model is defined by the first generative model information;
configuration is adopted. here,
- The first generative model information is information relating to the first data generative model, and is information reflecting a second data generative model different from the first data generative model.

According to the above configuration, by using the first data generation model in which the second generative model information about the second generative model is reflected, the other generative model learning used for learning the second generative model It is possible to generate diverse data corresponding to both the generative model training data used for training the first generative model and the second generative model without directly referring to the data for the generative model. can. Further, as an example, when the data generated in this exemplary embodiment is used for learning an inference model, the diversity of data for learning an inference model can be increased, and an inference capable of handling a variety of data with higher accuracy can be obtained. model can be obtained.

[Exemplary embodiment 3]
A third exemplary embodiment of the invention will now be described in detail with reference to the drawings. The information processing device 30 according to the present exemplary embodiment is a device that refers to generative model information that respectively defines a plurality of data generative models and generates updated generative model information.

The information processing device 30 according to this exemplary embodiment will be described in more detail below.

<Configuration of information processing device>
The configuration of the information processing device 30 according to this exemplary embodiment will be described with reference to FIG. FIG. 5 is a block diagram showing the configuration of the information processing device 30. As shown in FIG. As shown in FIG. 5 , the information processing device 30 includes a generative model information acquisition unit 301 and a generative model information update unit 302 .

Here, the generative model information acquisition unit 301 is an example of the generative model information acquisition means according to this exemplary embodiment. Also, the generative model information updating unit 302 is an example of the generative model information updating means according to this exemplary embodiment.

(generative model information acquisition part)
The generative model information acquisition unit 301 acquires first generative model information regarding a first data generation model and second generative model information regarding a second data generation model different from the first data generation model.

(First data generation model and second data generation model)
A first data generation model is a model that generates data according to input data. The first data generation model is learned in an information processing device different from the information processing device 30 using the data for generative model learning. The second data generation model is a model different from the first data generation model, and is a model that generates data according to input data. The second data generation model is learned using the generative model learning data in an information processing device different from the information processing device 30 and the information processing device that learns the first data generation model. The generative model learning data used for learning the first data generation model and the generative model learning data used for learning the second data generation model are, for example, acquired independently by each information processing device. data and can differ from each other.

(First generative model information and second generative model information)
The first generative model information is information relating to the first data generative model, and includes, for example, various parameters defining the first data generative model. The second generative model information is information relating to the first data generative model, and includes, for example, various parameters defining the second data generative model.

Note that the first generative model information in this exemplary embodiment does not necessarily reflect the second generative model information, unlike the first generative model information in

exemplary embodiments

1 and 2. .

(Generative model information update section)
The generative model information updating unit 302 refers to the first generative model information and the second generative model information to generate updated generative model information.

(Generative model information after update)
The first generative model information and the second generative model information are reflected in the updated generative model information. For example, the first data generation model and the second data generation model may be newly defined with reference to updated generation model information.

<Flow of information processing method>
The information processing method S30 executed by the information processing apparatus 30 configured as described above will be described with reference to FIG. FIG. 6 is a flow diagram showing the flow of the information processing method S30. As shown in FIG. 6, the information processing method S30 includes steps S301 to S203.

(Step S301)
In step S301, the generative model information acquisition unit 301 acquires first generative model information and second generative model information. For example, the generative model information acquisition unit 301 acquires first generative model information from an information processing apparatus that has learned the first data generative model. Also, for example, the generative model information acquisition unit 301 acquires second generative model information from the information processing apparatus that has learned the second data generative model.

(Step S302)
In step S302, the generative model information update unit 302 refers to the first generative model information and the second generative model information to generate post-update generative model information. As an example, the generative model information updating unit 302 may include information obtained by taking a weighted average of the first generative model information and the second generative model information in the updated generative model information.

<Effects of this exemplary embodiment>
As described above, in the information processing device 30 according to this exemplary embodiment,
Acquiring first generative model information about a first data generative model and second generative model information about a second data generative model different from the first data generative model;
generating updated generative model information by referring to the first generative model information and the second generative model information;
configuration is adopted.

According to the above configuration, the data generation model defined by referring to the updated generative model information includes (i) generative model learning data used for learning the first data generation model, and (ii) It functions as a data generation model learned by indirectly using the generative model learning data used for learning the second data generation model.

As a result, even in situations where acquisition of at least one of the data for generative model training in (i) and (ii) above should be avoided, it is possible to indirectly use both to generate more diverse data. data generation model can be generated. In other words, it is possible to generate a data generation model capable of generating diverse data as in the case of learning by directly using both (i) and (ii). In addition, as an example, when data generated using such a data generation model is used for training an inference model, the diversity of the data for training the inference model can be increased, resulting in more accurate and diverse data. It is possible to obtain a corresponding inference model.

Note that the information processing apparatus 30 according to this exemplary embodiment functions as a server in federated learning as an example, but this exemplary embodiment is not limited by the wording.

[Exemplary embodiment 4]
A fourth exemplary embodiment of the invention will now be described in detail with reference to the drawings. Components having the same functions as those described in exemplary embodiments 1 to 3 are denoted by the same reference numerals, and description thereof will not be repeated.

<Configuration of information processing system>
The configuration of an information processing system 1a according to this exemplary embodiment will be described with reference to FIG. FIG. 7 is a block diagram showing the configuration of the information processing system 1a. As shown in FIG. 7, the information processing system 1a includes a first information processing device 10a and a server 30a. The first information processing device 10a and the server 30a are connected via a network N. FIG. Specific examples of the network N include a wireless LAN (Local Area Network), a wired LAN, a WAN, a public line network, a mobile data communication network, or a combination of these networks. However, the configuration of network N is not limited to this.

(Configuration of first information processing device)
The first information processing device 10 a includes a learning data acquisition unit 101 , a generative model information acquisition unit 102 , a data generation model learning unit 103 and a transmission unit 110 .

The learning data acquisition unit 101 and the data generation model learning unit 103 are configured in the same manner as in the first exemplary embodiment.

The generative model information acquisition unit 102 is configured in substantially the same manner as in the first exemplary embodiment. However, this exemplary embodiment differs in that the first generative model information is obtained by receiving it from the server 30a.

The transmission unit 110 transmits third generative model information related to the first data generation model learned by the data generation model learning unit 103 to the server 30a.

(Third generative model information)
The third generative model information is information about the learned first data generation model, and includes, for example, parameters that define the learned first data generation model.

(Fourth generative model information)
The fourth generative model information is information about the second data generation model after learning, and includes, for example, parameters that define the second data generation model after learning.

(Server configuration)
The server 30 a includes a generative model information acquisition unit 301 , a generative model information update unit 302 and a transmission unit 310 .

(generative model information acquisition unit)
The generative model information acquisition unit 301 is configured in substantially the same manner as in the third exemplary embodiment. However, the present exemplary embodiment differs in that third generative model information and fourth generative model information are obtained instead of the first generative model information and the second generative model information.

(Generative model information update part)
The generative model information update unit 302 is configured almost similarly to the third exemplary embodiment. However, in this exemplary embodiment, the generative model information updater 302 refers to the third generative model information and the fourth generative model instead of referring to the first generative model information and the second generative model information. Generate updated generative model information by referring to the information.

The transmission unit 310 transmits the updated generative model information to the first information processing device 10a.

<Flow of information processing method>
An information processing method S1a executed by the information processing system 1a configured as described above will be described with reference to FIG. FIG. 8 is a flowchart showing the flow of the information processing method S1a. As shown in FIG. 8, the information processing method S1a includes S101a to S104a executed by the first information processing apparatus 10a and S301a to S303a executed by the server 30a. The information processing system 1a sequentially executes the information processing method S1a. The information processing method S1a to be executed for the n-th time will be described below. (n = 1, 2, ...)
(Step S101a)
In step S101a, the learning data acquisition unit 101 acquires generative model learning data.

(Step S102a)
In step S102a, the generative model information acquisition unit 102 acquires the first generative model information by receiving it from the server 30a. When the information processing method S1a is executed for the first time (n=1), the first generative model information received from the server 30a does not necessarily have to reflect the second generative model information. Further, when the information processing method S1a is executed for the second time or later (n≧2), the generative model information acquisition unit 102 acquires the The “updated generative model information” is acquired as the first generative model information. In this case, the acquired first generative model information (that is, the previous updated generative model information) is information that reflects the second generative model information (that is, the previous fourth generative model information). .

(Step S103a)
In step S103a, the data generation model learning unit 103 creates a first data generation model defined by referring to the first generative model information acquired by the generative model information acquisition unit 102, and converts the data for generative model learning into teacher data. used as a learning method. For example, the data generation model learning unit 103 uses the data for generative model learning to execute update processing for updating the parameter group defining the first data generation model. The data generation model learning unit 103 repeats the update process until the learning converges or until the specified number of times is reached. Thereby, the first data generation model is learned.

(Step S104a)
In step S104a, the transmission unit 110 transmits the third generative model information regarding the learned first data generative model to the server 30a.

(Step S301a)
In step S301a, the generative model information acquisition unit 301 of the server 30a acquires the third generative model information by receiving it from the first information processing apparatus 10a. Further, the generative model information acquisition unit 301 acquires the fourth generative model information by receiving it from another information processing apparatus that sequentially learns the second data generative model. Other information processing devices are devices different from the first information processing device 10a and the server 30a.

(Step S302a)
In step S302a, the generative model information update unit 302 refers to the third generative model information and the fourth generative model information to generate post-update generative model information. As an example, the generative model information updating unit 302 may include information obtained by taking a weighted average of the third generative model information and the fourth generative model information in the updated generative model information.

(Step S303a)
In step S303a, the transmission unit 310 transmits the updated generative model information to the first information processing apparatus 10a. The "updated generative model information" transmitted in this step is received by the first information processing apparatus 10a as "first generative model information" in step S102a in the next (n+1)th information processing method S1a. .

<Effects of this exemplary embodiment>
As described above, in the information processing system 1a according to this exemplary embodiment,
The first information processing device 10a
・ Acquire learning data (generative model learning data) of the first data generation model,
- first generative model information that is updated generative model information received from the server 30a and reflects second generative model information related to a second data generative model different from the first data generative model; , obtained as first generative model information about the first data generative model,
・Learning a first data generation model defined by referring to the acquired first generative model information using learning data (generative model learning data),
- Sending the third generative model information about the learned first data generative model to the server 30a;
configuration is adopted.

Moreover, the server 30a
Acquiring third generative model information and fourth generative model information related to the second data generative model after learning,
- Generate updated generative model information by referring to the third generative model information and the fourth generative model information;
- Transmitting the updated generative model information to the first information processing device 10a;
configuration is adopted.

According to the above configuration, the first information processing device 10a can acquire the first data generation model defined by referring to the updated generation model information received from the server 30a as a learning target. can. As a result, the first information processing device 10a does not directly acquire the other generative model learning data used by the other information processing device that generated the second generative model. The learning data can be used indirectly to generate a first data generation model capable of generating more diverse data. In other words, the first information processing device 10a learns by directly using both the generative model learning data acquired by the first information processing device and the generative model learning data acquired by another information processing device. A first data generation model can be generated that can similarly generate diverse data. In addition, as an example, when the data generated by such a first data generation model is used for learning an inference model, the diversity of the inference model learning data can be increased, and more accurate and diverse data can be obtained. It is possible to obtain an inference model that can correspond to

[Exemplary embodiment 5]
A fifth exemplary embodiment of the present invention will now be described in detail with reference to the drawings. Components having the same functions as the components described in exemplary embodiments 1 to 4 are denoted by the same reference numerals, and description thereof will not be repeated.

<Configuration of information processing system>
The configuration of an information processing system 1b according to this exemplary embodiment will be described with reference to FIG. FIG. 9 is a block diagram showing the configuration of the information processing system 1b. The information processing system 1b is a system that eliminates the imbalance in the data group for inference model learning and trains the inference model using the resolved data group. The information processing system 1b generates a data generation model for generating data to be added to the data group by federated learning in order to eliminate the imbalance.

As shown in FIG. 9, the information processing system 1b includes a first information processing device 10b-1, a second information processing device 10b-2, and a server 30b. The first information processing device 10b-1, the second information processing device 10b-2, and the server 30b are connected via a network N. FIG.

(inference model)
In this exemplary embodiment, the inference model is a model that infers the label species of the data. The inference model outputs information indicating one of a plurality of label types when data is input. The first information processing device 10b-1 uses the learning data acquired by the device to learn the first inference model. The second information processing device 10b-2 uses the learning data acquired by the device to learn the second inference model. The first inference model and the second inference model are different from each other. Note that the configurations of the plurality of label types that can be indicated by the data output by the first inference model and the data output by the second inference model do not necessarily have to be the same. For example, the first inference model outputs data indicating any of the labels L1, L3, L4, . You may output the data shown. Hereinafter, when there is no need to distinguish between the first inference model and the second inference model, they will simply be referred to as "inference model".

(Learning data, unbalanced data group)
Learning data is a piece of data acquired for use in learning an inference model, and is any of a plurality of pieces of data included in a data group. Each data piece included in the data group is labeled with one of a plurality of label types, and the percentage of data pieces labeled with a certain label type is less than or equal to a predetermined percentage. Hereinafter, such a data group is also referred to as an "unbalanced data group", and "a certain label species" is also referred to as a "minority label species".

In other words, the imbalanced data group includes a plurality of learning data (data pieces) each labeled with one of a plurality of label types, and at least one of the plurality of label types satisfies the minority label type condition. . For example, the condition for the minority label type may be, as described above, that the ratio of data pieces to which the label type is attached is equal to or less than a predetermined ratio, or that the data pieces to which the label type is attached may be less than the number of other label types. However, the conditions for minority label types are not limited to those described above.

The first imbalance data group acquired by the first information processing device 10b-1 and the second imbalance data group used by the second information processing device 10b-2 are at least partially different from each other. Here, the minority label species in the first imbalanced data group and the minority label species in the second imbalanced data group may or may not be the same. Incidentally, when there is no need to specifically distinguish between the first imbalance data group and the second imbalance data group, they will also simply be referred to as "imbalance data group".

A specific example when the minority label species in the first imbalanced data group and the second imbalanced data group are not the same will be described. A specific example of a data piece included in such an unbalanced data group is a product image. In this case, the first imbalance data group and the second imbalance data group are sets of product images obtained at different production plants. Each product image is labeled with one of "normal product", "defective product pattern A", and "defective product pattern B". In the first imbalanced data set, the percentage of product images labeled "Defective Pattern A" is less than or equal to a predetermined percentage. Or, the number of product images labeled with "defective product pattern A" is less than the number of product images labeled with "normal product", and the label with "defective product pattern B" is attached less than the number of product images shown. In this case, the minority label type in the first imbalance data group is "defective product pattern A." On the other hand, in the second imbalance data group, the percentage of product images labeled with "defective product pattern B" is less than or equal to the predetermined percentage. Or, the number of product images labeled with "defective product pattern B" is less than the number of product images labeled with "normal product", and the label with "defective product pattern A" is attached less than the number of product images shown. In this case, the minority label type in the second imbalance data group is "defective product pattern B".

Also, a specific example in which the minority label species in the first imbalanced data group and the second imbalanced data group are the same will be described. A specific example of a piece of data included in such an imbalanced data group is a piece of communication data in a financial system or communication system. In this case, the first imbalanced data set and the second imbalanced data set are sets of pieces of communication data obtained by different systems. There is an example in which each piece of communication data is labeled as "normal" or "abnormal", with "abnormal" being a minority label type. Another specific example of the data piece is a medical image. In this case, the first unbalanced data group and the second unbalanced data group are sets of medical images obtained at different medical institutions. Each medical image is labeled as "without lesion" or "with lesion", with "with lesion" being a minority label type. However, specific examples of data pieces and minority label types are not limited to those described above.

(Equilibrium data group)
The equilibrium data group includes a plurality of learning data (data pieces) each labeled with one of a plurality of label types, and none of the label types satisfies the minority label type condition. For example, in the balanced data group, the ratio of learning data with each label type is greater than or equal to the processing ratio, or the number of learning data with each label type is equal.

The first balanced data group generated by the first information processing device 10b-1 and the second balanced data group generated by the second information processing device 10b-2 are at least partially different from each other. However, if there is no need to specifically distinguish between the first equilibrium data group and the second equilibrium data group, they will simply be referred to as the "equilibrium data group".

(data generation model)
The data generation model generates data to be used as inference model learning data. Although the first information processing device 10b-1 generates a first data generation model and the second information processing device 10b-2 generates a second data generation model, they will be described separately. If it is not necessary, it will also be simply described as "data generation model". The data generation model is learned to generate learning data that can be teacher data for at least a few label types.

(Generative model learning data)
The generative model learning data is data for learning the data generative model. Part or all of the imbalance data group is used as generative model learning data.

(Inference model learning data)
The inference model learning data is data for learning the inference model. In this exemplary embodiment, a balanced data group in which the imbalance in the imbalanced data group is eliminated is used as inference model training data.

(Configuration of first information processing device)
First information processing device 10b-1 includes control unit 150-1 and communication unit 160-1. The control unit 150-1 includes a learning data acquisition unit 101-1, a generative model information acquisition unit 102-1, a data generation model learning unit 103-1, a seed data acquisition unit 201-1, and a data generation unit 202. -1, an inference model learning unit 203-1, an inference unit 204-1, and a communication unit 160-1.

(learning data acquisition unit)
The learning data acquisition unit 101-1 acquires a first imbalance data group. The details of the first imbalance data group are as described above.

(generative model information acquisition part)
The generative model information acquisition unit 102-1 is configured similarly to the generative model information acquisition unit 102 in the first exemplary embodiment.

(Data generation model learning part)
Data generation model learning section 103-1 learns the first data generation model using at least part of the first imbalance data group. For example, the data generation model learning unit 103-1 may learn at least one of an autoencoder including the first data generation model and a generative adversarial network including the first data generation model. Details of the first data generation model will be described later.

(Seed data acquisition part)
The seed data acquisition unit 201-1 acquires seed data to be input to the first data generation model. For example, the seed data acquisition unit 201-1 may acquire by generating seed data, or may be generated by another functional block (not shown) included in the device or another device (not shown). You may obtain seed data that has been

(data generator)
The data generator 202-1 generates data using seed data. Specifically, the data generation unit 202-1 generates learning data that can be teacher data for a small number of labels by inputting seed data into the first data generation model. For example, the data generation unit 202-1 inputs seed data and conditions for identifying minority label types into the first data generation model to generate learning data that can be teacher data for minority label types. good too. The data generator 202-1 also adds the generated learning data to the first imbalance data group to generate the first equilibrium data group.

(Inference model learning part)
The inference model learning unit 203-1 refers to the data generated by the data generation unit 202-1 to learn the first inference model. Specifically, inference model learning section 203-1 uses the first equilibrium data group as inference model learning data.

(reasoning part)
The inference unit 204-1 makes an inference using the inference model generated by the inference unit 204-1.

(communication department)
The communication unit 160-1 transmits to the server 30b third generative model information regarding the first data generative model learned by the data generative model learning unit 103-1.

(Configuration of second information processing device)
Second information processing device 10b-2 includes control unit 150-2 and communication unit 160-2. Control unit 150-2 includes learning data acquisition unit 101-2, generation model information acquisition unit 102-2, data generation model learning unit 103-2, seed data acquisition unit 201-2, and data generation unit 202. -2, an inference model learning unit 203-2, and an inference unit 204-2.

For details of the configuration of the second information processing device 10b-2, in the description of the configuration of the first information processing device 10b-1, the suffixes “1” and “2” of each reference numeral are read interchangeably, and “second 1” and “second” are read interchangeably, and “third” and “fourth” are interchanged.

(Server configuration)
Server 30 b includes a control unit 350 and a communication unit 360 . The control unit 350 includes a generative model information acquisition unit 301 , a generative model information update unit 302 and a communication unit 360 .

(generative model information acquisition unit)
The generative model information acquisition unit 301 is configured similarly to the generative model information acquisition unit 301 in the fourth exemplary embodiment.

(Generative model information update section)
The generative model information update unit 302 is configured similarly to the generative model information update unit 302 in the fourth exemplary embodiment. Specifically, in this exemplary embodiment, the generative model information updater 302 generates updated generative model information by averaging the third generative model information and the fourth generative model information.

(communication department)
The communication unit 360 distributes the updated generative model information to the first information processing device 10b-1 and the second information processing device 10b-2.

<Flow of information processing method for generating data generation model>
An information processing method S1b executed by the information processing system 1b configured as described above will be described with reference to FIG. FIG. 10 is a flowchart for explaining the flow of the information processing method S1b. The information processing system 1b generates a data generation model by sequentially executing the information processing method S1b. In FIG. 10, the reference numerals including (n) at the end indicate the steps of the information processing method S1b to be executed for the nth time, and the reference numerals including (n+1) at the end are the information processing to be executed for the n+1th time. 4 shows the steps of method S1b.

(Step S101-1)
In the n-th step S101-1, the learning data acquisition unit 101-1 of the first information processing device 10b-1 acquires the first imbalance data group.

(Step S360)
In the n-th step S360, the communication unit 360 of the server 30b distributes the generative model information to the first information processing device 10b-1 and the second information processing device 10b-2. When n=1, predetermined information is applied as the generative model information to be distributed. If n≧2, the “updated generative model information” updated in step S302 for the n−1 time is applied as the generative model information to be distributed. Step S302 will be described later.

(Step S102-1)
In the n-th step S102-1, the generative model information acquisition unit 102-1 of the first information processing device 10b-1 acquires the generative model information delivered from the server 30b as first generative model information.

(Step S103-1)
In the n-th step S103-1, the data generation model learning unit 103-1 of the first information processing device 10b-1 refers to the first generative model information acquired by the generative model information acquisition unit 102-1. A defined first data generation model is trained. For learning, part or all of the learning data included in the first imbalance data group is used as teacher data. Specifically, the data generation model learning unit 103-1 uses the learning data included in the first imbalance data group to determine whether the first data generation model can be training data of at least a minority label type. The update process for updating the parameter group so as to output the data is repeated until learning converges or until a specified number of times is reached.

(Step S160-1)
In the n-th step S160-1, the communication unit 160-1 of the first information processing device 10b-1 transmits the third generative model information regarding the learned first data generative model to the server 30b.

(Steps S101-2 to S160-2)
The second information processing device 10b-2 executes steps S101-2, S102-2, S103-2, and S160-2 for the n-th time. Details of these steps are described in the descriptions of steps S101-1, S102-1, S103-1, and S160-1 executed by the first information processing device 10b-1. The same description is given by replacing "2" with each other, "first" and "second" with each other, and "third" and "fourth" with each other.

(Step S301)
In step S301 of the n+1 time, the generative model information acquisition unit 301 of the server 30b acquires the third generative model information by receiving it from the first information processing device 10b-1. Further, the generative model information acquisition unit 301 acquires the fourth generative model information by receiving it from the second information processing device 10b-2.

(Step S302)
In the n+1th step S302, the generative model information updating unit 302 refers to the third generative model information and the fourth generative model information to generate post-update generative model information. Specifically, the generative model information update unit 302 includes information obtained by taking a weighted average of the third generative model information and the fourth generative model information in the updated generative model information.

(Step S360)
In the n+1-th step S360, the communication unit 360 distributes the updated generative model information to the first information processing device 10b-1 and the second information processing device 10b-2.

Also, the first information processing device 10b-1 executes steps S101-1, S102-1, S103-1, and S160-1 for the n+1th time. The second information processing device 10b-2 executes steps S101-2, S102-2, S103-2, and S160-2 for the (n+1)th time. The details are as described in each n-th step.

By sequentially executing the information processing method S1b in this way, the first information processing device 10b-1 directly acquires the learning data acquired by the second information processing device 10b-2. can be used indirectly without As a result, the first information processing device 10b-1 can generate a first data generation model capable of generating more diverse data. Also, the second information processing device 10b-2 can indirectly use the learning data acquired by the first information processing device 10b-1 without directly acquiring it. As a result, the second information processing device 10b-2 can generate a second data generation model capable of generating more diverse data.

<Flow of information processing method for generating inference model>
The information processing method S20b executed by the first information processing device 10b-1 and the second information processing device 10b-2 will be described with reference to FIG. FIG. 11 is a flowchart for explaining the flow of the information processing method S20b. The first information processing device 10b-1 generates a first inference model by executing the information processing method S20b. The second information processing device 10b-2 generates a second inference model by executing the information processing method S20b. An example in which the first information processing device 10b-1 executes the information processing method S20b will be described below. As for the example in which the second information processing device 10b-2 executes the information processing method S20b, in the following description, the suffixes “1” and “2” of each reference numeral are replaced with each other, and “first” and “second 2” are interchanged with each other. As shown in FIG. 11, the information processing method S20b includes steps S201 to S203.

(Step S201)
In step S201, the seed data acquisition unit 201-1 acquires seed data. Specifically, the seed data acquisition unit 201-1 acquires at least as many pieces of seed data as can eliminate the imbalance in the first imbalance data group. When obtaining a plurality of seed data, it is desirable that these seed data are different from each other. The number of imbalances that can be resolved is, for example, the number of learning data with a minority label in the first imbalance data group, and a predetermined percentage in the first equilibrium data group after the imbalance is resolved. is the difference between the number of For example, the number of learning data to which minority labels are attached in the first imbalanced data group is 20, the number of learning data contained in the first imbalanced data group is 100, and the predetermined ratio is 0.3. In this case, the possible number to resolve the imbalance is fifteen. In addition, the number of imbalances that can be resolved is, for example, the number of learning data with minority label types and the number of learning data with other label types in the first imbalance data group. is the difference between For example, assume that there are two kinds of label types, the number of learning data to which minority label types are attached is 20, and the number of learning data to which other label types are attached is 40. In this case, the possible number to resolve the imbalance is twenty. Note that the number of imbalances that can be resolved is not limited to the example described above.

(Step S202)
In step S202, the data generation unit 202-1 inputs the seed data generated in step S201 to the first data generation model, thereby generating learning data that can be teacher data for a minority label type. Further, the data generator 202-1 adds the generated learning data to the first imbalance data group. As a result, a first balanced data group in which the imbalance in the first imbalanced data group is eliminated is generated.

(Step S203)
In step S203, the inference model learning unit 203-1 learns the first inference model. For learning, each learning data included in the first equilibrium data group is used as inference model learning data.

<Specific example of data generation model>
Here, specific examples of the first data generation model and the second data generation model will be described. In the following description, the first data generation model and the second data generation model are simply referred to as "data generation model" unless it is necessary to distinguish them from each other. In addition, when it is not necessary to specifically distinguish which information processing device includes each functional block included in the first information processing device 10b-1 and the second information processing device 10b-2, "-1" or "-2" at the end of the reference number is omitted.

(Specific example 1)
A specific example 1 in which the data generation model is implemented as a decoder included in the variational autoencoder will be described with reference to FIG. FIG. 12 is a schematic diagram illustrating specific example 1 of the data generation model. As shown in FIG. 12, the variational autoencoder trained by the data generation model learning unit 103 includes an encoder EN1 and a decoder DE1. Encoder EN1 outputs latent variable Z when data X is input. Decoder DE1 outputs data X' when latent variable Z is input. The data generation model learning unit 103 inputs the data X to which the minority label type is attached among the learning data X acquired by the learning data acquisition unit 101 to the encoder EN1, and the data X′ and the data X output by the decoder DE1. Learning is performed so that the difference between Specifically, the data generation model learning unit 103 uses the learning data X to which the minority label seed is attached to set a parameter group that defines the encoder EN1 and the decoder DE1 so that the difference between X and X' becomes smaller. Execute the update process to update. In addition, the data generation model learning unit 103 repeats the update process until the learning converges or until it reaches a specified number of times. The decoder DE1 after learning becomes a data generation model that generates learning data that can be teacher data for a minority label type.

The seed data acquisition unit 201 generates seed data S to be input to the decoder DE1 after learning. Specifically, the seed data acquisition unit 201 generates, as the seed data S, the data in the range indicated by the latent variable Z that is output from the learning data X to which the minority label seed is attached. For example, the seed data acquisition unit 201 may generate the seed data S by randomly sampling data in the range indicated by the latent variable Z. FIG.

By inputting the seed data S to the decoder DE2, the data generation unit 202 generates learning data that can be teacher data for a small number of labels.

(Specific example 2)
A specific example 2 in which the data generation model is implemented as a decoder included in the conditional variational autoencoder will be described with reference to FIG. FIG. 13 is a schematic diagram illustrating a specific example 2 of the data generation model. As shown in FIG. 13, the conditional variational autoencoder trained by the data generation model learning unit 103 includes an encoder EN2 and a decoder DE2. Encoder EN2 outputs latent variable Z when data X and condition Y are input. Decoder DE1 outputs data X' when latent variable Z and condition Y are input. Here, as the condition Y, information specifying the label type is used.

The data generation model learning unit 103 inputs the learning data X acquired by the learning data acquisition unit 101 to the encoder EN2, and sends the condition Y specifying the label type attached to the learning data to the encoder EN2 and the decoder DE2. input. Further, the data generation model learning unit 103 performs learning so that the difference between the data X' output by the decoder DE2 and the data X becomes smaller. Specifically, the data generation model learning unit 103 uses the learning data X acquired by the learning data acquisition unit 101 to reduce the difference between X and X′ by using parameters that define the encoder EN2 and the decoder DE2. Execute update processing to update the group. In addition, the data generation model learning unit 103 repeats the update process until the learning converges or until it reaches a specified number of times. The decoder DE2 after learning becomes a data generation model that generates learning data that can be teacher data of the label type specified by the condition Y. FIG.

The seed data acquisition unit 201 generates seed data S to be input to the post-learning decoder DE2. Specifically, the seed data acquisition unit 201 refers to the latent variable Z output by inputting the learning data and the condition Y for identifying the minority label type to the encoder EN, and determines the range indicated by the latent variable Z. data is generated as seed data S; For example, the seed data acquisition unit 201 may generate the seed data S by randomly sampling data in the range indicated by the latent variable Z. FIG.

The data generation unit 202 generates learning data that can be teacher data for minority label types by inputting seed data S and a condition Y for specifying minority label types to the decoder DE2.

(Specific example 3)
A specific example 3 in which the data generation model is implemented as a generator included in the adversarial generation network will be described with reference to FIG. FIG. 14 is a schematic diagram illustrating a specific example 3 of the data generation model. As shown in FIG. 14, the generative adversarial network trained by the data generation model learning unit 103 includes a generator GEN and a discriminator DIS. The generator GEN outputs data GD when noise ND is input. The discriminator DIS determines whether the data GD is the true data X when the data GD and the true data X are inputted.

The data generation model learning unit 103 inputs the noise ND to the generator GEN, and inputs the data X to which the minority label type is attached among the learning data X acquired by the learning data acquisition unit 101 to the discriminator DIS. . Also, the data generation model learning unit 103 performs learning so that the data GD generated by the generator GEN is determined to be true. Specifically, the data generation model learning unit 103 uses the learning data X to which the minority label seed is attached, and sets the parameter group that defines the generator GEN and the discriminator DIS so that the data GD is determined to be true. Execute the update process to update the . In addition, the data generation model learning unit 103 repeats the update process until the learning converges or until it reaches a specified number of times. The generator GEN after learning becomes a data generation model that generates learning data that can be teacher data for minority label types.

The seed data acquisition unit 201 generates seed data S to be input to the post-learning generator GEN. Specifically, the seed data acquisition unit 201 may generate randomly generated noise as the seed data S. FIG.

The data generation unit 202 generates learning data that can be teacher data for a small number of labels by inputting the seed data S to the post-learning generator GEN.

(Other specific examples)
Another example is implementing the data generation model as a generator included in a conditional generative adversarial network. In such a specific example, the condition Y is further input to the generator GEN and discriminator DIS shown in FIG. The generator GEN outputs data GD when condition Y and noise ND are input. The discriminator DIS determines whether or not the data GD is the true data X when the condition Y, the data GD and the true data X are input.

In this case, the data generation model learning unit 103 inputs any of the learning data X acquired by the learning data acquiring unit 101 and the condition Y specifying the label type attached to the data X to the discriminator DIS. . Further, the data generation model learning unit 103 inputs the condition Y specifying the label type and the noise ND to the generator GEN. Also, the data generation model learning unit 103 performs learning so that the data GD generated by the generator GEN is determined to be true. Specifically, the data generation model learning unit 103 uses the learning data X acquired by the learning data acquisition unit 101 to define the generator GEN and the discriminator DIS so that the data GD is determined to be true. Execute update processing to update the parameter group. In addition, the data generation model learning unit 103 repeats the update process until the learning converges or until it reaches a specified number of times. The generator GEN after learning becomes a data generation model that generates learning data that can be teacher data of the label type specified by the condition Y.

The seed data acquisition unit 201 generates seed data S to be input to the post-learning generator GEN. The data generation unit 202 generates learning data that can be teacher data for minority label types by inputting seed data S and conditions Y for specifying minority label types to the decoder DE2.

<Effects of this exemplary embodiment>
As described above, in this exemplary embodiment,
Each of the first information processing device 10b-1 and the second information processing device 10b-2 acquires an imbalanced data group, and performs associative learning using the acquired imbalanced data group to perform minority-labeled teacher training. Generating a data generation model for generating training data that can be data,
configuration is adopted.

According to the above configuration, even if the number of learning data with minority label types is not sufficient, it is possible to indirectly use the minority label type learning data acquired by other information processing devices. As a result, it is possible to generate a data generation model capable of generating more diverse data as learning data that can be teacher data for a small number of label types.

In addition, according to the above configuration, by adding learning data that can be teacher data for a minority label type generated using a data generation model, a balanced data group in which the imbalance in the imbalanced data group is eliminated is generated. be able to. As a result, by using such a balanced data group as inference model learning data, the inference model can be learned with higher accuracy.

[Exemplary embodiment 6]
A sixth exemplary embodiment of the invention will now be described in detail with reference to the drawings. Components having the same functions as the components described in the exemplary embodiments 1 to 5 are denoted by the same reference numerals, and the description thereof will not be repeated.

<Configuration of information processing system>
The configuration of an information processing system 1c according to this exemplary embodiment will be described with reference to FIG. FIG. 15 is a block diagram showing the configuration of the information processing system 1c. The information processing system 1c is a modification of the information processing system 1c according to exemplary embodiment 5 so as to generate an inference model by federated learning. That is, the information processing system 1c generates the inference model itself by federated learning in addition to generating the data generation model for eliminating the imbalance in the imbalanced data group used for learning the inference model by federated learning.

As shown in FIG. 15, the information processing system 1c includes a first information processing device 10c-1, a second information processing device 10c-2, and a server 30c. The first information processing device 10c-1, the second information processing device 10c-2, and the server 30c are connected via a network N. FIG.

(Configuration of first information processing device)
The first information processing device 10c-1 includes an inference model information acquisition unit 205-1 in addition to the same configuration as the first information processing device 10b-1 according to the fifth exemplary embodiment. Also, in this exemplary embodiment, the configurations of the inference model learning unit 203-1 and the communication unit 160-1 are slightly different from those in the fifth exemplary embodiment. Other functional blocks are the same as those of exemplary embodiment 5, so detailed description will not be repeated.

(Inference model information acquisition unit)
The inference model information acquisition unit 205-1 acquires first inference model information regarding the first inference model.

(First inference model information)
The first inference model information includes, for example, various parameters that define the first inference model. In the exemplary embodiment, the first inference model information reflects second inference model information about a second inference model that is different from the first inference model.

(Inference model learning part)
The inference model learning unit 203-1 learns the first inference model defined by referring to the first inference model information using the data generated by the data generation unit 202-1. Specifically, the inference model learning unit 203-1 adds the data generated by the data generation unit 202-1 to the first balanced data group in which the imbalance in the first imbalanced data group is eliminated. is used for training the first inference model.

(communication department)
The communication unit 160-1, in addition to being configured in the same manner as in exemplary embodiment 5, transmits the third inference model information to the server 30c.

(Third inference model information)
The third inference model information is information about the first inference model after learning, and includes, for example, parameters that define the first inference model after learning.

(Fourth inference model information)
The fourth inference model information is information about the second inference model after learning, and includes, for example, parameters that define the second inference model after learning.

(Configuration of second information processing device)
The second information processing device 10c-2 includes an inference model information acquisition unit 205-2 in addition to the same configuration as the second information processing device 10b-2 according to the fifth exemplary embodiment. For details of the configuration of the second information processing device 10c-2, in the description of the configuration of the first information processing device 10c-1, "1" and "2" at the end of each reference numeral are replaced with each other, 1” and “second” are read interchangeably, and “third” and “fourth” are interchanged.

(Server configuration)
The server 30c includes an inference model information acquisition unit 303 and an inference model information update unit 304 in addition to the same configuration as the server 30b according to the fifth exemplary embodiment. Also, the configuration of the communication unit 360 is slightly different from that of the fifth exemplary embodiment. Other functional blocks are the same as those of exemplary embodiment 5, so detailed description will not be repeated.

(Inference model information acquisition unit)
The inference model information acquisition unit 303 acquires third inference model information regarding the first inference model and fourth inference model information regarding the second inference model.

(Inference model information update unit)
The inference model information update unit 304 generates updated inference model information by referring to the third inference model information and the fourth inference model information.

(Updated inference model information)
The updated inference model information reflects the third generative model information and the fourth generative model information.

(communication department)
In addition to being configured in the same manner as in Exemplary Embodiment 5, the communication unit 360 transmits updated inference model information to the first information processing device 10c-1 and the second information processing device 10c-2. To deliver.

<Flow of information processing method for generating data generation model>
The information processing system 1c configured as described above sequentially executes the information processing method S1b shown in FIG. Generate a data generation model for

<Flow of information processing method for generating inference model>
An information processing method S1c executed by the information processing system 1c will be described with reference to FIG. FIG. 16 is a flowchart for explaining the flow of the information processing method S1c. The information processing system 1c generates an inference model by sequentially executing the information processing method S1c. In FIG. 16 , reference symbols including (n) at the end indicate steps of the information processing method S1c to be executed for the nth time, and reference symbols including (n+1) at the end are information processing to be performed for the n+1th time. 4 shows the steps of method S1c.

(Step S201-1)
In the n-th step S201-1, the seed data acquisition unit 201-1 of the first information processing device 10c-1 acquires seed data by the number capable of resolving the imbalance in the first imbalance data group. .

(Step S202-1)
In the n-th step S202-1, the data generation unit 202-1 of the first information processing device 10c-1 inputs each seed data to the first data generation model, thereby becoming teacher data of a few label types. Generate training data to obtain. As a result, the number of learning data that can eliminate the imbalance is generated. Further, the data generator 202-1 adds the generated learning data to the first imbalance data group. This produces a first set of balanced data.

(Step S370)
In the n-th step S370, the communication unit 360 of the server 30c distributes the updated inference model information to the first information processing device 10c-1 and the second information processing device 10c-2. When n=1, predetermined information is applied as the inference model information to be distributed. If n≧2, the “updated inference model information” updated in step S304 for the (n−1)th time is applied as the inference model information to be distributed. Step S304 will be described later.

(Step S205-1)
In the n-th step S205-1, the inference model information acquisition unit 205-1 of the first information processing device 10c-1 acquires the inference model information delivered from the server 30c as the first inference model information.

(Step S203-1)
In the n-th step S203-1, the inference model learning unit 203-1 of the first information processing device 10c-1 refers to the first inference model information acquired by the inference model information acquisition unit 205-1 to define A first inference model is trained. The first equilibrium data group generated in step S202-1 is used for learning. Specifically, the data generation model learning unit 103-1 performs update processing for updating the parameter group of the first inference model using the learning data included in the first equilibrium data group until learning converges. , or repeat until a specified number of times is reached.

(Step S170-1)
In the n-th step S170-1, the communication unit 160-1 of the first information processing device 10c-1 transmits the third inference model information regarding the learned first inference model to the server 30c.

(Steps S201-2, S202-2, S205-2, S203-2, S170-2)
The second information processing device 10c-2 executes steps S201-2, S202-2, S205-2, S203-2, and S170-2 for the n-th time. Details of these steps are described in the description of steps S201-1, S202-1, S205-1, S203-1, and S170-1 executed by the first information processing device 10c-1. The same description is given by replacing "1" and "2" with each other, "first" and "second" with each other, and "third" and "fourth" with each other.

(Step S303)
In step S303 of the (n+1)th time, the inference model information acquisition unit 303 of the server 30c acquires the third inference model information by receiving it from the first information processing device 10c-1. Also, the inference model information acquisition unit 303 acquires the fourth inference model information by receiving it from the second information processing device 10c-2.

(Step S304)
In the n+1-th step S304, the inference model information updating unit 304 refers to the third inference model information and the fourth inference model information to generate updated inference model information. Specifically, the inference model information updating unit 304 includes information obtained by taking a weighted average of the third inference model information and the fourth inference model information in the updated inference model information.

(Step S370)
In the n+1-th step S370, the communication unit 360 distributes the updated inference model information to the first information processing device 10c-1 and the second information processing device 10c-2.

Also, the first information processing device 10c-1 executes steps S201-1, S202-1, S205-1, S203-1, and S170-1 for the (n+1)th time. The second information processing device 10c-2 executes steps S201-2, S202-2, S205-2, S203-2, and S170-2 for the (n+1)th time. The details are as described in each n-th step.

In this way, by sequentially executing the information processing method S1c, the first information processing device 10c-1 directly acquires or generates learning data obtained by the second information processing device 10c-2. It can be used indirectly without obtaining it, and the first inference model can be generated accurately and efficiently. Further, the second information processing device 10c-2 can indirectly use the learning data acquired or generated by the first information processing device 10c-1 without directly acquiring it, and the second inference model can be generated more accurately and efficiently.

<Effects of this exemplary embodiment>
As described above, in this exemplary embodiment,
Each of the first information processing device 10c-1 and the second information processing device 10c-2 generates an inference model using a balanced data group in which imbalance is eliminated using a data generation model generated by federated learning. Let them learn associatively.

According to the above configuration, even if the number of learning data included in the balanced data group whose imbalance is resolved using the data generation model is not sufficient, the balanced data group generated by the other information processing device The included training data can be used indirectly. As a result, even in situations where direct acquisition of equilibrium data groups generated by other information processing devices should be avoided, inference models can be generated efficiently due to the increased amount of learning data that can be referenced indirectly. be able to.

[Exemplary Embodiment 7]
A seventh exemplary embodiment of the present invention will now be described in detail with reference to the drawings. Components having the same functions as the components described in the exemplary embodiments 1 to 6 are denoted by the same reference numerals, and the description thereof will not be repeated.

<Configuration of information processing system>
A configuration of an information processing system 1d according to this exemplary embodiment will be described with reference to FIG. FIG. 17 is a block diagram showing the configuration of the information processing system 1d. The information processing system 1d is a modification of the information processing system 1c according to the sixth exemplary embodiment so as not to include the server 30c. That is, the information processing system 1d performs federated learning of the data generation model and the inference model without using the server.

As shown in FIG. 17, the information processing system 1d includes a first information processing device 10d-1 and a second information processing device 10d-2. A first information processing device 10d-1 and a second information processing device 10d-2 are connected via a network N. FIG.

(Configuration of first information processing device)
The first information processing device 10d-1 includes a model information updating unit 260-1 in addition to the same configuration as the first information processing device 10c-1 according to the sixth exemplary embodiment. Further, in this exemplary embodiment, the configurations of the generative model information acquisition unit 102-1, the data generation model learning unit 103-1, the inference model information acquisition unit 205-1, and the inference model learning unit 203-1 are It is slightly different from the sixth embodiment. Other functional blocks are the same as those in the sixth exemplary embodiment, so detailed description will not be repeated.

(generative model information acquisition unit)
The generative model information acquisition unit 102-1 acquires fourth generative model information from the second information processing device 10d-2. The fourth generative model information is information about the second data generative model after learning which is learned by the second information processing device 10d-2.

(Inference model information acquisition unit)
The inference model information acquisition unit 205-1 acquires fourth inference model information from the second information processing device 10d-2. The fourth inference model information is information about the second inference model after learning which is learned by the second information processing device 10d-2.

(Model information update section)
The model information updating unit 206-1 generates updated generative model information by referring to the third generative model information and the fourth generative model information. Here, the third generative model information is information about the learned first data generative model learned by the data generative model learning unit 103-1.

Also, the model information update unit 206 generates updated inference model information by referring to the third inference model information and the fourth inference model information. Here, the third inference model information is information about the first inference model after learning which the inference model learning unit 203-1 has made to learn.

(Data generation model learning part)
The data generation model learning unit 103-1 learns the first data generation model defined by the updated generation model information generated by the model information update unit 206-1 using the first imbalance data group. .

(Inference model learning part)
The inference model learning unit 203-1 uses the first equilibrium data group to learn the first inference model defined by the updated inference model information generated by the model information update unit 206-1.

(communication department)
The communication unit 160-1 transmits the third generative model information and the third inference model information to the second information processing device 10d-2.

(Configuration of second information processing device)
The second information processing device 10d-2 includes a model information updating unit 260-2 in addition to the same configuration as the second information processing device 10c-2 according to the sixth exemplary embodiment. For details of the configuration of the second information processing device 10d-2, in the description of the configuration of the first information processing device 10d-1, the suffixes “1” and “2” of each reference numeral are replaced with each other, and “second 1” and “second” are read interchangeably, and “third” and “fourth” are interchanged.

<Flow of information processing method for generating data generation model>
The information processing system 1d configured as described above modifies and sequentially executes the information processing method S1b shown in FIG. 10 to generate a first data generation model and a second data generation model. do. Specifically, the information processing method S1b is modified such that step S302 that was executed by the server 30b is executed by the first information processing device 10d-1 and the second information processing device 10d-2 instead. be. As an example, the first information processing apparatus 10d-1 executes step S302-1 between steps S102-1 and S103-1. The modified information processing method S1b will be described below.

(Step S101-1)
The n-th step S101-1 is the same as in the fifth exemplary embodiment.

(Step S102-1)
The n-th step S102-1 is executed when n≧2. In the n-th step S102-1, the generative model information acquisition unit 102-1 of the first information processing device 10d-1 acquires fourth generative model information from the second information processing device 10d-2. The fourth generative model information is transmitted from the second information processing device 10d-2 in the previous (n-1)th step S160-2. If n=1, execution of step S102-1 is omitted.

(Step S302-1)
In the n-th step S302-1, the model information updating unit 206-1 of the first information processing device 10d-1 refers to the third generative model information and the fourth generative model information, and updates the Generate generative model information for . When n≧2, the third generative model information is information about the first data generative model after learning in the previous (n−1)th step S103-1. When n=1, the model information updating unit 206-1 generates predetermined information as updated generative model information.

(Step S103-1)
In the n-th step S103-1, the data generation model learning unit 103-1 of the first information processing device 10d-1 converts the first data generation model defined by the updated generative model information into the first Learn using an imbalanced data group.

(Step S160-1)
In the n-th step S160-1, the communication unit 160-1 of the first information processing device 10d-1 transmits new third generative model information regarding the learned first data generative model to the second information. It is transmitted to the processing device 10d-2.

(Steps S101-2, S102-2, S302-2, S103-2, S160-2)
The second information processing device 10d-2 executes steps S101-2, S102-2, S103-2, and S160-2 for the n-th time. Details of these steps are described in the description of steps S101-1, S102-1, S302-1, S103-1, and S160-1 executed by the first information processing device 10d-1. The same description is given by replacing "1" and "2" with each other, "first" and "second" with each other, and "third" and "fourth" with each other.

<Flow of information processing method for generating inference model>
The information processing system 1d transforms the information processing method S1c shown in FIG. 16 and sequentially executes it to generate a first inference model and a second inference model. The information processing method S1c is modified such that step S304, which was executed by the server 30c, is executed by the first information processing device 10d-1 and the second information processing device 10d-2 instead. As an example, the first information processing apparatus 10d-1 executes step S304-1 between steps S205-1 and S203-1. The modified information processing method S1c will be described below.

(Steps S201-1 to S202-1)
The nth steps S201-1 to S202-1 are the same as in the sixth exemplary embodiment.

(Step S205-1)
The n-th step S205-1 is executed when n≧2. In the n-th step S205-1, the inference model information acquisition unit 205-1 of the first information processing device 10d-1 acquires the fourth inference model information from the second information processing device 10d-2. The fourth inference model information is transmitted from the second information processing device 10d-2 in the previous (n-1)th step S170-2. If n=1, execution of step S205-1 is omitted.

(Step S304-1)
In the n-th step S304-1, the model information updating unit 206-1 of the first information processing device 10d-1 refers to the third inference model information and the fourth inference model information, and updates the Generate inference model information for When n≧2, the third inference model information is information about the first inference model after learning in the previous (n−1)th step S203-1. When n=1, the model information updating unit 206-1 generates predetermined information as updated inference model information.

(Step S203-1)
In the n-th step S203-1, the inference model learning unit 203-1 of the first information processing device 10d-1 converts the first inference model defined by the updated inference model information to the first equilibrium data. Learn using a group.

(Step S170-1)
In the n-th step S170-1, the communication unit 160-1 of the first information processing device 10d-1 transmits new third inference model information regarding the learned first inference model to the second information processing unit. Send to device 10d-2.

(Steps S201-2, S202-2, S205-2, S304-2, S203-2, S170-2)
The second information processing device 10d-2 executes steps S201-2, S202-2, S205-2, S304-2, S203-2, and S170-2 for the n-th time. For details of these steps, refer to the descriptions of steps S201-1, S202-1, S205-1, S304-1, S203-1, and S170-1 executed by the first information processing device 10d-1. The same explanation can be made by replacing the suffixes "1" and "2" with each other, "first" and "second" with each other, and with "third" and "fourth" with each other.

<Effects of this exemplary embodiment>
As described above, in the first information processing device 10d-1 of this exemplary embodiment,
receiving fourth generative model information about the second data generative model after learning from another information processing device;
generating updated generative model information by referring to the third generative model information and the fourth generative model information regarding the learned first data generative model in the own device;
transmitting the third generative model information to another information processing device;
configuration is adopted.

According to the above configuration, the same effects as those of exemplary embodiments 5 and 6 can be obtained without requiring a server. Note that this exemplary embodiment acquires the fourth generative model information and the fourth inference model information from another information processing device. For this reason, the exemplary embodiment is suitable in situations where sharing of information possessed by multiple information processing devices is less problematic.

[Exemplary embodiment 8]
An eighth exemplary embodiment of the invention will now be described in detail with reference to the drawings. Components having the same functions as those described in exemplary embodiments 1 to 7 are denoted by the same reference numerals, and description thereof will not be repeated.

<Configuration of information processing system>
The configuration of an information processing system 1e according to this exemplary embodiment will be described with reference to FIG. FIG. 18 is a block diagram showing the configuration of the information processing system 1e. The information processing system 1e is a mode in which the functions of the server 30c in exemplary embodiment 6 are distributed and realized.

As shown in FIG. 18, the information processing system 1e includes a first information processing device 10e-1, a second information processing device 10e-2, a first server 30e, and a second server 40e. A first information processing device 10e-1, a second information processing device 10e-2, a first server 30e, and a second server 40e are connected via a network N. FIG.

(Configuration of first information processing device)
The first information processing device 10e-1 is configured substantially similarly to the first information processing device 10c-1 according to the sixth exemplary embodiment. However, in this exemplary embodiment, the configurations of the generative model information acquisition unit 102-1, the inference model information acquisition unit 205-1, and the communication unit 160-1 are slightly different from those in the sixth exemplary embodiment. Other functional blocks are the same as those in the sixth exemplary embodiment, so detailed description will not be repeated.

(generative model information acquisition part)
The generative model information acquisition unit 102-1 is configured substantially in the same manner as in the sixth exemplary embodiment, but differs in that the first generative model information to be acquired is distributed by the first server 30e.

(Inference model information acquisition unit)
The inference model information acquisition unit 205-1 is configured substantially in the same manner as in the sixth exemplary embodiment, but differs in that the first inference model information to be acquired is distributed by the second server 40e.

(communication department)
The communication unit 160-1 is configured in substantially the same manner as in the sixth exemplary embodiment, except that the destination of the third generative model information is the first server 30e. Another difference is that the destination of the third inference model information is the second server 40e.

(Configuration of second information processing device)
The second information processing device 10e-2 is configured substantially similarly to the second information processing device 10c-2 according to the sixth exemplary embodiment. For details of the configuration of the second information processing device 10e-2, in the description of the configuration of the first information processing device 10e-1, "1" and "2" at the end of each reference numeral are replaced with each other, 1” and “second” are read interchangeably, and “third” and “fourth” are interchanged. However, "first server" and "second server" are not read interchangeably.

(Configuration of first server)
The first server 30e functions as a server in federated learning of data generation models. The first server 30e has a generative model information acquisition unit 301, a generative model information update unit 302, and a communication unit 360 among the functional blocks included in the server 30c according to the sixth exemplary embodiment. The details of each of these functional blocks are the same as in the sixth exemplary embodiment.

(Configuration of the second server)
The second server 40e functions as a server in federated learning of inference models. The second server 40e is a device different from the first server 30e. The second server 40e has an inference model information acquisition unit 303, an inference model information update unit 304, and a communication unit 460 among the functional blocks included in the server 30c according to the sixth exemplary embodiment. The details of each of these functional blocks are the same as in the sixth exemplary embodiment.

<Flow of information processing method for generating data generation model>
The information processing system 1e configured as described above sequentially executes the information processing method S1b shown in FIG. Generate a data generation model for However, in this exemplary embodiment, the steps performed by the server 30b in the information processing method S1b are performed by the first server 30e.

<Flow of information processing method for generating inference model>
The information processing system 1e configured as described above sequentially executes the information processing method S1c shown in FIG. Generate an inference model. However, in this exemplary embodiment, the steps performed by the server 30c in the information processing method S1c are performed by the second server 40e.

<Effects of this exemplary embodiment>
As described above, in this exemplary embodiment,
A server that performs federated learning of a data generation model and a server that performs federated learning of an inference model using a balanced data group whose imbalance has been resolved are realized by different devices,
configuration is adopted.

According to the above configuration, it is possible to distribute the load of the servers required to achieve the same effects as in the sixth exemplary embodiment. Also, a device suitable for each load can be applied as each server. For example, a plurality of information processing devices including a first information processing device 10e-1 and a second information processing device 10e-2 may participate in federated learning of data generation models and federated learning of inference models. be done. In such a case, there is a possibility that the number of information processing devices participating in the federated learning of the data generation model and the number of information processing devices participating in the federated learning of the inference model are significantly different. In this case, it is assumed that the loads on the first server 30e and the second server 40e are significantly different. In such a case, devices suitable for respective loads can be applied as the first server 30e and the second server 40e.

[Exemplary embodiment 9]
A ninth exemplary embodiment of the present invention will now be described in detail with reference to the drawings. Components having the same functions as the components described in the exemplary embodiments 1 to 8 are denoted by the same reference numerals, and the description thereof will not be repeated.

<Configuration of information processing system>
The configuration of an information processing system 1f according to this exemplary embodiment will be described with reference to FIG. FIG. 19 is a block diagram showing the configuration of the information processing system 1f. The information processing system 1f is a mode in which the functions of the server 30b in the information processing system 1b according to the fifth exemplary embodiment are implemented in a distributed manner. Here, in exemplary embodiment 5, between federated learning clients (the first information processing device 10b-1 and the second information processing device 10b-2), each other information processing device obtains and generates This is an example of a form that takes into consideration situations in which direct reference to information should be avoided. By modifying this, the exemplary embodiment realizes an example of a form that further considers the situation in which the server should avoid directly referring to the information acquired and generated by each client in federated learning.

As shown in FIG. 19, the information processing system 1f includes a first information processing device 10f-1, a second information processing device 10f-2, a first server 30f-1, a second server 30f-2, and It includes a third server 30f-3. The first information processing device 10f-1, the second information processing device 10f-2, the first server 30f-1, the second server 30f-2, and the third server 30f-3 are connected via network N;

(data piece of generative model information)
In this exemplary embodiment, federated learning clients (first information processing device 10f-1 and second information processing device 10f-2) and servers (first server 30f-1 and second server 30f) -2, and the third server 30f-3), it transmits and receives a piece of data generated from the generative model information.

Here, it is assumed that three data pieces A, B, and C are generated from the generative model information, but the number of data pieces is not limited to this. The data piece A, the data piece B, and the data piece C are information different from each other, and are information for generating generative model information. Data piece A, data piece B, and data piece C cannot restore the generative model information by referring to either one or two of them, but they are information that can generate generative model information by referring to all of them. be. For example, the data pieces A to C may be data pieces in which the generative model information is randomly divided into three so that the total sum does not change.

(Overview of associative learning using data pieces)
Each client generates data pieces A to C from the generative model information related to the generative model learned by its own device, and distributes and transmits them to each server. Since each server does not receive all of data piece A, data piece B, and data piece C from one client, it cannot directly refer to the original generative model information. Each server integrates data piece A (or data piece B, data piece C) received from each client to generate integrated data piece A (or data piece B, data piece C). Each client can acquire all of the integrated data piece A, the integrated data piece B, and the integrated data piece C from two or more servers. Each client can refer to the post-integration data piece A, the post-integration data piece B, and the post-integration data piece C to restore the updated generative model information.

(Configuration of first information processing device)
The first information processing device 10f-1 is configured almost similarly to the first information processing device 10b-1 according to the fifth exemplary embodiment. However, in this exemplary embodiment, the configurations of the generative model information acquisition unit 102-1 and the communication unit 160-1 are slightly different from those in the fifth exemplary embodiment. Other functional blocks are the same as those of exemplary embodiment 5, so detailed description will not be repeated.

(communication department)
Communication unit 160-1 generates third data pieces A to C from the third generative model information regarding the updated first data generative model. Also, the communication unit 160-1 transmits the third data pieces A and B to the first server 30f-1. Also, the communication unit 160-1 transmits the third data pieces B and C to the second server 30f-2. Also, the communication unit 160-1 transmits the third data pieces C and A to the third server 30f-3. The third data pieces A to C are applied as "first data pieces A to C" in each server.

(generative model information acquisition unit)
The generative model information acquisition unit 102-1 transfers the integrated data piece A, the integrated data piece B, and the integrated data piece C to the first server 30f-1, the second server 30f-2, and the Obtained by receiving from at least two of the third servers 30f-3. In addition, the generative model information acquisition unit 102-1 refers to the data piece A after integration, the data piece B after integration, and the data piece C after integration, and obtains the first generative model information after update. to generate Here, the data piece A after integration is the sum of the first data piece A and the second data piece A. FIG. Also, the data piece B after integration is the sum of the first data piece B and the second data piece B. FIG. Also, the data piece C after integration is the sum of the first data piece C and the second data piece C. As shown in FIG. The generative model information acquisition unit 102-1 divides the total sum of the integrated data piece A, the integrated data piece B, and the integrated data piece C by the total number of the information processing devices 10 (here, 2). obtains the average of the sum of the first data pieces A to C (ie, the first generative model information) and the sum of the second data pieces A to C (ie, the second generative model information). Further, the generative model information acquiring unit 102-1 uses the average of these as the updated first generative model information. The second data pieces A to C (that is, the second generative model information) are reflected in the updated first generative model information.

(Configuration of second information processing device)
The second information processing device 10f-2 is configured almost similarly to the second information processing device 10b-2 according to the fifth exemplary embodiment. For details of the differences from exemplary embodiment 5, in the description of the configuration of the first information processing device 10f-1, the suffixes “1” and “2” of each reference numeral are replaced with each other, and “first” and A similar description will be made by replacing "second" with each other and "third" and "fourth" with each other. However, "first server", "second server" and "third server" are not interchangeable.

(Configuration of first server)
The first server 30f-1 functions as one of a plurality of distributed servers in federated learning of data generation models. First server 30f-1 includes control unit 350-1 and communication unit 360-1. Control unit 350-1 includes a generative model information acquisition unit 301-1 and a generative model information update unit 302-1.

(generative model information acquisition part)
The generative model information acquisition unit 301-1 receives the first data pieces A and B from the first information processing device 10f-1. Also, the generative model information acquisition unit 301-1 receives the second data pieces A and B from the second information processing device 10f-2.

(Generative model information update unit)
The generative model information updating unit 302-1 integrates the first data piece A and the second data piece A to generate the data piece A after integration. Further, the generative model information updating unit 302-1 integrates the first data piece B and the second data piece B to generate the data piece B after integration. As described above, here, as the data piece A (or data piece B) after integration, the first data piece A (or data piece B) and the second data piece A (or data piece B ).

(communication department)
The communication unit 360-1 receives a request for the integrated data piece A and the integrated data piece B from one or both of the first information processing device 10f-1 and the second information processing device 10f-2. Deliver according to.

(Configuration of the second server)
The second server 30f-2 functions as one of a plurality of distributed servers in federated learning of data generation models. The second server 30f-2 includes a control section 350-2 and a communication section 360-2. Control unit 350-2 includes a generative model information acquisition unit 301-2 and a generative model information update unit 302-2.

(generative model information acquisition part)
The generative model information acquisition unit 301-2 receives the first data pieces B and C from the first information processing device 10f-1. Also, the generative model information acquisition unit 301-2 receives the second data pieces B and C from the second information processing device 10f-2.

(Generative model information update unit)
The generative model information updating unit 302-2 integrates the first data piece B and the second data piece B to generate the data piece B after integration. Further, the generative model information updating unit 302-2 integrates the first data piece C and the second data piece C to generate the data piece C after integration. As described above, here, as the data piece B (or data piece C) after integration, the first data piece B (or data piece C) and the second data piece B (or data piece C ).

(communication department)
The communication unit 360-2 receives a request for the integrated data piece B and the integrated data piece C from one or both of the first information processing device 10f-1 and the second information processing device 10f-2. Deliver according to.

(Configuration of the third server)
The third server 30f-3 functions as one of a plurality of distributed servers in federated learning of data generation models. The third server 30f-3 includes a control section 350-3 and a communication section 360-3. Control unit 350-3 includes a generative model information acquisition unit 301-3 and a generative model information update unit 302-3.

(generative model information acquisition part)
The generative model information acquisition unit 301-3 receives the first data pieces C and A from the first information processing device 10f-1. Also, the generative model information acquisition unit 301-3 receives the second data pieces C and A from the second information processing device 10f-2.

(Generative model information update section)
The generative model information updating unit 302-3 refers to the first data piece C and the second data piece C to generate the data piece C after integration. Further, the generative model information update unit 302-3 integrates the first data piece A and the second data piece A to generate the data piece A after integration. As described above, here, as the data piece C (or data piece A) after integration, the first data piece C (or data piece A) and the second data piece C (or data piece A ).

(communication department)
The communication unit 360-3 receives a request for the integrated data piece C and the integrated data piece A from one or both of the first information processing device 10f-1 and the second information processing device 10f-2. Deliver according to.

<Flow of information processing method for generating data generation model>
The information processing system 1f configured as described above modifies the information processing method S1b shown in FIG. 10 and sequentially executes it to generate a first data generation model and a second data generation model. do.

Specifically, the information processing method S1b causes the first server 30f-1, the second server 30f-2, and the third server 30f-3 to execute steps S301, S302, and S360, respectively. Transformed. The operations of the first server 30f-1, the second server 30f-2, and the third server 30f-3 in these steps are described in the description of steps S301, S302, and S360 in FIG. 3 generative model information” should be read as “first data piece”, “fourth generative model information” should be read as “second data piece”, and “updated generative model information” should be read as “data after integration The same explanation is given by reading "piece". Further, in the information processing method S1b, the operation of step S102-1 is modified so as to generate the first generative model information by referring to the data pieces A to C after integration. Further, the operation of step S160-1 is to transfer two each of the third data pieces A to C generated from the third generative model information to different servers (first server 30f-1, second server 30f). -2, and modified to send to a third server 30f-3). Further, the operation of step S160-2 is to send two each of the fourth data pieces A to C generated from the fourth generative model information to different servers (first server 30f-1, second server 30f). -2, and modified to send to a third server 30f-3).

<Effects of this exemplary embodiment>
As described above, in this exemplary embodiment,
A server that performs federated learning of data generation models is distributed to a plurality of different devices,
configuration is adopted.

Here, for example, due to privacy considerations, there are cases where it is difficult for the server itself in federated learning to acquire information held by each information processing device. In this case, it may not be preferable for the server to directly acquire the third generative model information and the fourth generative model information. According to the above configuration, the first server 30f-1, the second server 30f-2, and the third server 30f-3 provide information about the first generative model after learning and the second generative model after learning. There is no direct acquisition of information about the generative model of On the other hand, in the first information processing device 10f-1 and the second information processing device 10f-2, the information about the first generative model after learning and the information about the second generative model after learning are reflected. The updated generative model information can be generated by referring to the information received from each server. Therefore, an effect similar to that of exemplary embodiment 5 can be obtained.

Note that in this exemplary embodiment, an example has been described in which the functions of the server 30b in exemplary embodiment 5 are implemented by distributing them to three servers. The functions of the server 30b are not limited to this, and may be implemented by being distributed to four or more servers. In that case, the first information processing device 10f-1 and the second information processing device 10f-2 refer to the generative model information about the data generative model after learning, respectively, to obtain four or more data pieces A, B, C, .

Note that the method of dividing and restoring the generative model information performed by the information processing system 1f according to this exemplary embodiment is also called multi-party computation as an example, but this exemplary embodiment is limited by this wording. not to be

[Exemplary embodiment 10]
A tenth exemplary embodiment of the present invention will now be described in detail with reference to the drawings. This exemplary embodiment includes a first information processing device (10b-1, 10c-1, 10d-1, 10e-1, 10f-1) according to any of exemplary embodiments 5-9, and a second 10b-2, 10c-2, 10d-2, 10e-2, 10f-2) in which the learning data acquisition units 101-1 and 101-2 are modified. In the following description, if there is no need to distinguish between the first information processing apparatus and the second information processing apparatus, "-1" and "-2" at the end of the reference numerals are omitted as appropriate. Describe. The learning data acquisition unit 101 in this exemplary embodiment will be described with reference to FIG. FIG. 20 is a diagram showing the configuration of the learning data acquisition unit 101. As shown in FIG. Other configurations and the flow of the information processing method are the same as those described in each exemplary embodiment, so detailed description thereof will be omitted.

(Learning data selection part)
As shown in FIG. 20 , the learning data acquisition section 101 includes a learning data selection section 111 . The learning data selection unit 111 is an example of selection means according to this exemplary embodiment. The learning data selection unit 111 selects an imbalanced data group from a plurality of data group candidates.

For example, the learning data selection unit 111 may select data that satisfies the condition of being an imbalanced data group among the data set A, the data set B, the data set C, . . . stored in the storage unit SR. good. For example, such a condition may be the minority label type condition described above. Note that the storage unit SR may be included in the

information processing apparatuses

10b, 10c, 10d, 10e, and 10f, or may be included in an external device.

<Effects of this exemplary embodiment>
As described above, in this exemplary embodiment,
selecting an imbalanced data group to be used for learning the data generation model from a plurality of data group candidates;
configuration is adopted.

According to the above configuration, each information processing device selects its own data set so that the imbalance is eliminated as a whole, so that the data generation model can be learned more stably.

[Modification]
Each exemplary embodiment described above has been described as including one or two information processing devices. Without limitation, each exemplary embodiment can be modified to include more than two information processors. Thereby, each information processing device can generate a data generation model by indirectly using learning data used in a larger number of other information processing devices without directly acquiring the data.

Further, in the sixth to ninth exemplary embodiments described above, the functional blocks included in each of the two information processing devices (the first information processing device and the second information processing device) are dispersed and included in a plurality of devices. may be For example, the functional blocks included in the first information processing device (or the second information processing device) are: (i) functional blocks related to data generation model learning (learning data acquisition unit, generation model information acquisition unit, and data generation model learning unit); (ii) a device including functional blocks related to data generation (seed data acquisition unit and data generation unit); and (iii) functional blocks related to inference model learning (inference model information acquisition unit and inference model learning unit) and (iv) an apparatus including a function block (inference unit) related to inference.

[Example of realization by software]

Information processing apparatus

10, 10a, 10b-1, 10b-2, 10c-1, 10c-2, 10d-1, 10d-2, 10e-1, 10e-2, 10f-1, 10f-2, 20,

server

30, 30a, 30b, 30c, 30d, 30e, 30f, 40e, 50f (hereinafter referred to as each information processing device 10 to 20 and each server 30 to 50) are partially or entirely functionally integrated circuits (IC It may be implemented by hardware such as a chip), or may be implemented by software.

In the latter case, each of the information processing devices 10 to 20 and each of the servers 30 to 50 described above is implemented by, for example, a computer that executes program instructions that are software that implements each function. An example of such a computer (hereinafter referred to as computer C) is shown in FIG. Computer C comprises at least one processor C1 and at least one memory C2. A program P for operating the computer C as each of the information processing apparatuses 10 to 20 and each of the servers 30 to 50 is recorded in the memory C2. In the computer C, the processor C1 reads the program P from the memory C2 and executes it, thereby implementing the functions of the information processing devices 10-20 and the servers 30-50.

As the processor C1, for example, CPU (Central Processing Unit), GPU (Graphic Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating point number Processing Unit), PPU (Physics Processing Unit) , a microcontroller, or a combination thereof. As the memory C2, for example, a flash memory, HDD (Hard Disk Drive), SSD (Solid State Drive), or a combination thereof can be used.

Note that the computer C may further include a RAM (Random Access Memory) for expanding the program P during execution and temporarily storing various data. Computer C may further include a communication interface for sending and receiving data to and from other devices. Computer C may further include an input/output interface for connecting input/output devices such as a keyboard, mouse, display, and printer.

In addition, the program P can be recorded on a non-temporary tangible recording medium M that is readable by the computer C. As such a recording medium M, for example, a tape, disk, card, semiconductor memory, programmable logic circuit, or the like can be used. The computer C can acquire the program P via such a recording medium M. Also, the program P can be transmitted via a transmission medium. As such a transmission medium, for example, a communication network or broadcast waves can be used. Computer C can also obtain program P via such a transmission medium.

[Appendix 1]
The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. For example, embodiments obtained by appropriately combining the technical means disclosed in the embodiments described above are also included in the technical scope of the present invention.

[Appendix 2]
Some or all of the above-described embodiments may also be described as follows. However, the present invention is not limited to the embodiments described below.

(Appendix 1)
learning data acquisition means for acquiring learning data;
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model Generative model information acquisition means for acquiring
data generation model learning means for learning the first data generation model defined by referring to the first generation model information using the learning data;
Information processing device equipped with.

According to the above configuration, the first data generation model is obtained by indirectly using the other learning data used to generate the second data generation model in the other information processing apparatus without directly obtaining the data. to learn As a result, it is possible to generate a first data generation model capable of generating more diverse data even in a situation where acquisition of other learning data used in other information processing devices should be avoided. can be done.

(Appendix 2)
The information processing apparatus according to supplementary note 1, further comprising transmitting means for transmitting third generative model information related to the first data generative model learned by the data generative model learning means.

According to the above configuration, the third data generation model information regarding the first data generation model after learning can be provided to other devices. As a result, the learning data used for learning the first data generation model can be indirectly used for learning without being directly acquired by another device.

(Appendix 3)
The generative model information acquisition means includes:
Acquiring fourth generative model information about the second data generative model after learning,
The information processing device is
third generative model information about the first data generative model learned by the data generative model learning means;
The information processing apparatus according to appendix 1, further comprising generative model information updating means for generating updated generative model information with reference to the fourth generative model information.

According to the above configuration, by referring to the fourth generative model information, it is possible to indirectly use other learning data used to generate the second data generative model in another information processing apparatus. . As a result, it is possible to generate a first data generation model capable of generating more diverse data even in a situation where acquisition of other learning data used in other information processing devices should be avoided. can be done.

(Appendix 4)
The learning data acquisition means is
A data group containing a plurality of data pieces, each data piece labeled with one of a plurality of label types, and included in the data group in which the ratio of data pieces labeled with a certain label type is equal to or less than a predetermined ratio 4. The information processing apparatus according to any one of appendices 1 to 3, wherein a piece of data is acquired as the learning data.

According to the above configuration, the first data generation model is trained using an imbalanced data group with an imbalanced number of labeled data pieces as learning data. As a result, the first data generation model can be used to generate data for resolving imbalances in such data sets.

(Appendix 5)
The learning data acquisition means is
The information processing apparatus according to appendix 4, further comprising selection means for selecting the data group from a plurality of data group candidates.

According to the above configuration, by selecting the imbalanced data group, the effect described in appendix 4 can be achieved more reliably.

(Appendix 6)
The data generation model learning means is
an autoencoder including the first data generation model; and
6. The information processing apparatus according to any one of appendices 1 to 5, wherein at least one of generative adversarial networks including the first data generation model is learned.

According to the above configuration, it is possible to realize the first data generation model for achieving the above effects.

(Appendix 7)
The autoencoder is
7. The information processing device according to appendix 6, comprising at least one of a variational autoencoder and a conditional variational autoencoder.

(Appendix 8)
data generation means for generating data using the first data generation model;
8. The information processing apparatus according to any one of appendices 1 to 7, further comprising inference model learning means for learning the first inference model with reference to the data generated by the data generation means. .

According to the above configuration, the data generated by the first data generation model having the above-described effect is referred to, so the first inference model can be learned with high accuracy.

(Appendix 9)
Acquiring first inference model information relating to the first inference model, the first inference model information reflecting second inference model information relating to a second inference model different from the first inference model. further comprising inference model information acquisition means for
The inference model learning means learns the first inference model defined by referring to the first inference model information using the data generated by the data generation means. The information processing device described.

According to the above configuration, the first inference model is learned by indirectly using the other learning data used to generate the second inference model in the other information processing device without directly obtaining the data. Let As a result, the first inference model can be learned with high accuracy even in a situation where acquisition of other learning data used in another information processing apparatus should be avoided.

(Appendix 10)
seed data acquisition means for acquiring seed data;
Data generation means for generating data using the seed data,
The data generation means is
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model An information processing apparatus characterized by generating data using a first data generation model defined with reference to.

According to the above configuration, it is possible to generate more diverse data by using the first data generation model learned by reflecting the second generative model information regarding the second generative model.

(Appendix 11)
11. The information processing apparatus according to appendix 10, further comprising inference model learning means for learning the first inference model with reference to the data generated by the data generation means.

(Appendix 12)
Acquiring first inference model information relating to the first inference model, the first inference model information reflecting second inference model information relating to a second inference model different from the first inference model. further comprising inference model information acquisition means for
The inference model learning means learns the first inference model defined by referring to the first inference model information using the data generated by the data generation means. The information processing device described.

(Appendix 13)
Generative model information acquisition means for acquiring first generative model information regarding a first data generation model and second generative model information regarding a second data generation model different from the first data generation model;
Generative model information updating means for generating updated generative model information by referring to the first generative model information and the second generative model information;
An information processing device comprising:

According to the above configuration, (i) the learning data used for learning the first data generation model and (ii) the learning data used for learning the second data generation model are indirectly can generate a data generation model trained using As a result, even in a situation where acquisition of at least one of (i) and (ii) data for learning should be avoided, the data generation model can be trained to generate more diverse data. .

(Appendix 14)
transmitting means for transmitting the updated generative model information to a first information processing device implementing the first data generation model and a second information processing device implementing the second data generation model; 14. The information processing apparatus according to appendix 13, further comprising:

According to the above configuration, even if the first information processing device and the second information processing device should avoid providing each other with the learning data used by both, A data generation model learned by indirectly using learning data can be provided to the first information processing device and the second information processing device.

(Appendix 15)
The generative model information updating means includes:
15. The information processing apparatus according to appendix 13 or 14, wherein the updated generative model information is generated by averaging the first generative model information and the second generative model information.

According to the above configuration, learning is realized by indirectly using the learning data used for learning the first data generation model and the learning data used for learning the second data generation model. .

(Appendix 16)
An information processing system including a first information processing device and a server,
The first information processing device is
learning data acquisition means for acquiring learning data;
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model a generative model information acquisition means for acquiring from the server;
data generation model learning means for learning the first data generation model defined by referring to the first generation model information using learning data acquired by the learning data acquisition means;
transmitting means for transmitting third generative model information related to the first data generative model learned by the data generative model learning means;
The server is
the third generative model information; and
a generative model information acquiring means for acquiring fourth generative model information about the second data generative model after learning;
Generative model information updating means for generating updated generative model information by referring to the third generative model information and the fourth generative model information;
and transmitting means for transmitting the updated generative model information to the first information processing apparatus.

According to the above configuration, it is possible to realize a system that has the same effect as Appendix 1 by using the information processing device and the server.

(Appendix 17)
obtaining training data;
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model and
An information processing apparatus method, comprising: learning the first data generation model defined by referring to the first generation model information, using the learning data.

According to the above configuration, the same effect as Appendix 1 can be obtained.

(Appendix 18)
obtaining seed data; and generating data using the seed data;
including
In the step of generating the data,
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model An information processing method characterized by generating data using a first data generation model defined with reference to.

According to the above configuration, the same effects as in Supplementary Note 10 are obtained.

(Appendix 19)
obtaining first generative model information about a first data generative model and second generative model information about a second data generative model different from the first data generative model; and
generating updated generative model information by referring to the first generative model information and the second generative model information;
An information processing method comprising:

　According to the above configuration, the same effects as in Supplementary Note 13 are obtained.

(Appendix 20)
A program for causing a computer to function as an information processing device, the computer comprising:
learning data acquisition means for acquiring learning data;
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model Generative model information acquisition means for acquiring
data generation model learning means for learning the first data generation model defined by referring to the first generation model information using the learning data;
A program that acts as a

(Appendix 21)
A program for causing a computer to function as an information processing device, the computer comprising:
seed data acquisition means for acquiring seed data;
Data generation means for generating data using the seed data, the first generation model information relating to the first data generation model, the second data generation model different from the first data generation model A program functioning as data generating means for generating data using a first data generation model defined by referring to first generative model information in which second generative model information regarding is reflected.

(Appendix 22)
A program for causing a computer to function as an information processing device, the computer comprising:
Generative model information acquisition means for acquiring first generative model information regarding a first data generation model and second generative model information regarding a second data generation model different from the first data generation model;
Generative model information updating means for generating updated generative model information by referring to the first generative model information and the second generative model information;
A program that acts as a

[Appendix 3]
Some or all of the embodiments described above can also be expressed as follows.

(Appendix 23)
at least one processor, said processor comprising:
learning data acquisition processing for acquiring learning data;
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model Generative model information acquisition processing for acquiring
and a data generation model learning process of learning the first data generation model defined by referring to the first generation model information using the learning data.

The information processing apparatus may further include a memory, in which the learning data acquisition process, the generative model information acquisition process, and the data generative model learning process are executed by the processor. A program may be stored for causing the Also, this program may be recorded in a computer-readable non-temporary tangible recording medium.

(Appendix 24)
at least one processor, said processor comprising:
a seed data acquisition process for acquiring seed data;
A data generation process for generating data using the seed data, the first generation model information relating to the first data generation model, and a second data generation model different from the first data generation model and a data generation process for generating data using a first data generation model defined by referring to first generation model information in which second generation model information regarding is reflected.

The information processing apparatus may further include a memory, and the memory may store a program for causing the processor to execute the seed data acquisition process and the data generation process. . Also, this program may be recorded in a computer-readable non-temporary tangible recording medium.

(Appendix 25)
at least one processor, said processor comprising:
Generative model information acquisition processing for acquiring first generative model information related to a first data generation model and second generative model information related to a second data generation model different from the first data generation model;
and a generative model information update process for generating updated generative model information by referring to the first generative model information and the second generative model information.

The information processing apparatus may further include a memory, and the memory stores a program for causing the processor to execute the generative model information acquisition process and the generative model information update process. may Also, this program may be recorded in a computer-readable non-temporary tangible recording medium.

1a to 1f

information processing systems

10, 10a to 10f, 20

information processing devices

30, 30a to 30f, 40e, 50f information processing devices (servers)
101 learning

data acquisition units

102, 301, 501 generative model information acquisition unit 103 data generation

model learning units

160, 360, 460, 560

communication units

110, 310 transmission unit 111 learning data selection units 150, 350 control unit 201 seed Data acquisition unit 202 Data generation unit 203 Inference model learning unit 204 Inference units 205, 303 Inference model information acquisition unit 206 Model information update units 302, 502 Generative model information update unit 304 Inference model information update unit C Computer C1 Processor C2 Memory

Claims

learning data acquisition means for acquiring learning data;
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model Generative model information acquisition means for acquiring
data generation model learning means for learning the first data generation model defined by referring to the first generation model information using the learning data;
Information processing device equipped with.
2. The information processing apparatus according to claim 1, further comprising transmitting means for transmitting third generative model information related to said first data generative model learned by said data generative model learning means.
The generative model information acquisition means includes:
Acquiring fourth generative model information about the second data generative model after learning,
The information processing device is
third generative model information about the first data generative model learned by the data generative model learning means;
2. The information processing apparatus according to claim 1, further comprising generative model information updating means for generating updated generative model information with reference to said fourth generative model information.
The learning data acquisition means is
A data group containing a plurality of data pieces, each data piece labeled with one of a plurality of label types, and included in the data group in which the ratio of data pieces labeled with a certain label type is equal to or less than a predetermined ratio 4. The information processing apparatus according to any one of claims 1 to 3, wherein a piece of data is obtained as said learning data.
The learning data acquisition means is
5. An information processing apparatus according to claim 4, further comprising selection means for selecting said data group from a plurality of data group candidates.
The data generation model learning means is
an autoencoder including the first data generation model; and
6. The information processing apparatus according to any one of claims 1 to 5, wherein at least one of generative adversarial networks including the first data generation model is learned.
The autoencoder is
7. An information processing apparatus according to claim 6, comprising at least one of a variational autoencoder and a conditional variational autoencoder.
data generation means for generating data using the first data generation model;
8. The information processing according to claim 1, further comprising inference model learning means for learning the first inference model with reference to the data generated by said data generation means. Device.
Acquiring first inference model information relating to the first inference model, the first inference model information reflecting second inference model information relating to a second inference model different from the first inference model. further comprising inference model information acquisition means for
8. The inference model learning means learns the first inference model defined by referring to the first inference model information using the data generated by the data generation means. The information processing device according to .
seed data acquisition means for acquiring seed data;
Data generation means for generating data using the seed data,
The data generation means is
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model An information processing apparatus characterized by generating data using a first data generation model defined with reference to.
11. An information processing apparatus according to claim 10, further comprising inference model learning means for learning the first inference model with reference to the data generated by said data generation means.
Acquiring first inference model information relating to the first inference model, the first inference model information reflecting second inference model information relating to a second inference model different from the first inference model. further comprising inference model information acquisition means for
11. The inference model learning means learns the first inference model defined by referring to the first inference model information using the data generated by the data generation means. The information processing device according to .
Generative model information acquisition means for acquiring first generative model information regarding a first data generation model and second generative model information regarding a second data generation model different from the first data generation model;
Generative model information updating means for generating updated generative model information by referring to the first generative model information and the second generative model information;
An information processing device comprising:
transmitting means for transmitting the updated generative model information to a first information processing device implementing the first data generation model and a second information processing device implementing the second data generation model; 14. The information processing apparatus according to claim 13, further comprising:
The generative model information updating means includes:
15. The information processing apparatus according to claim 13, wherein the updated generative model information is generated by averaging the first generative model information and the second generative model information.
An information processing system including a first information processing device and a server,
The first information processing device is
learning data acquisition means for acquiring learning data;
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model a generative model information acquisition means for acquiring from the server;
data generation model learning means for learning the first data generation model defined by referring to the first generation model information using learning data acquired by the learning data acquisition means;
transmitting means for transmitting third generative model information related to the first data generative model learned by the data generative model learning means;
The server is
the third generative model information; and
a generative model information acquiring means for acquiring fourth generative model information about the second data generative model after learning;
Generative model information updating means for generating updated generative model information by referring to the third generative model information and the fourth generative model information;
and transmitting means for transmitting the updated generative model information to the first information processing apparatus.
obtaining training data;
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model and
An information processing apparatus method, comprising: learning the first data generation model defined by referring to the first generation model information, using the learning data.
obtaining seed data; and generating data using the seed data;
including
In the step of generating the data,
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model An information processing method characterized by generating data using a first data generation model defined with reference to.
obtaining first generative model information about a first data generative model and second generative model information about a second data generative model different from the first data generative model; and
generating updated generative model information by referring to the first generative model information and the second generative model information;
An information processing method comprising:
A program for causing a computer to function as an information processing device, the computer comprising:
learning data acquisition means for acquiring learning data;
First generative model information relating to a first data generation model, the first generative model information reflecting second generative model information relating to a second data generation model different from the first data generation model Generative model information acquisition means for acquiring
data generation model learning means for learning the first data generation model defined by referring to the first generation model information using the learning data;
A program that acts as a
A program for causing a computer to function as an information processing device, the computer comprising:
seed data acquisition means for acquiring seed data;
Data generation means for generating data using the seed data, the first generation model information relating to the first data generation model, the second data generation model different from the first data generation model A program functioning as data generating means for generating data using a first data generation model defined by referring to first generative model information in which second generative model information regarding is reflected.
A program for causing a computer to function as an information processing device, the computer comprising:
Generative model information acquisition means for acquiring first generative model information regarding a first data generation model and second generative model information regarding a second data generation model different from the first data generation model;
Generative model information updating means for generating updated generative model information by referring to the first generative model information and the second generative model information;
A program that acts as a