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

Abstract

An information processing system 100 comprises a database 10 of distributed type. The database 10 of distributed type stores verification information about information pertaining to a machine learning model. The verification information is transmitted from a first terminal device 30. The verification information is transmitted from the distributed database 10 to a second terminal device 40 in response to a request from the second terminal device 40.

Description

Technical Field

[0001] The present disclosure relates to a terminal device for executing distributed processing, an information processing system using these, and an information processing method, and further relates to a program for realizing these. Mu Related.

Background Art

[0002] In recent years, in the medical field, medical devices using artificial intelligence have been developed. Examples of medical devices using artificial intelligence include, for example, image diagnostic devices. An image diagnostic device is combined with an endoscope, an X-ray imaging device, etc., and detects a lesion by applying a captured image to a machine learning model (see, for example, Patent Documents 1 and 2). A machine learning model is constructed, for example, by using image data and teacher data as training data.

[0003] Medical devices need to be verified, approved, and authorized by a public institution before being sold. Examples of institutions (verifiers) that perform such verification include the Pharmaceuticals and Medical Devices Agency (hereinafter also referred to as "PMDA").

[0004] By the way, in a medical device using artificial intelligence, a verifier may request a medical device manufacturer to provide data used for constructing a machine learning model during the verification process. Disclosure of such data is disadvantageous to the medical device manufacturer from the viewpoint of leakage of know-how, etc. On the other hand, when a verifier approves without publicly releasing such data to the medical device manufacturer, it is desirable to prevent the medical device manufacturer from committing an illegal act of falsifying the data later.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

[0006] Therefore, a system is needed that allows for the exchange of data between medical device manufacturers and verifiers without disclosing the data used for approval and while preventing its tampering. However, such a system does not currently exist.

[0007] One example of the purpose of this disclosure is to provide a mechanism to prevent tampering with data used in the approval of medical devices without disclosing it. [Means for solving the problem]

[0008] To achieve the above objectives, the information processing system in one aspect of this disclosure is: Equipped with a distributed database, The aforementioned distributed database stores validation information regarding machine learning models, The verification information is transmitted from the first terminal device and, in response to a request from the second terminal device, is transmitted from the distributed database to the second terminal device. It is characterized by the following:

[0009] To achieve the above objective, the first terminal device in one aspect of this disclosure is A means for generating validation information from information about a machine learning model, A verification information transmission means that transmits and stores the calculated verification information in a distributed database, It is characterized by having the following features.

[0010] To achieve the above objective, the second terminal device in one aspect of this disclosure is: A means for obtaining validation information about machine learning models stored in a distributed database, A verification means for verifying a machine learning model by determining whether the verification information obtained from sources other than the distributed database is the same as the verification information obtained by the verification information acquisition means. Verification result transmission means for transmitting and storing the verification results from the verification unit to the distributed database, It is characterized by having the following features.

[0011] Furthermore, in order to achieve the above objectives, the information processing method in one aspect of this disclosure is: The first terminal device generates validation information from information about the machine learning model and transmits the calculated validation information to a distributed database. The distributed database stores the transmitted verification information and transmits the verification information to the second terminal device in response to a request from the second terminal device. The second terminal device is characterized by acquiring the verification information stored in the distributed database, determining whether the verification information obtained from sources other than the distributed database is the same as the verification information stored in the distributed database, and verifying the machine learning model.

[0012] To achieve the above objectives, the first aspect of this disclosure program teeth, On the computer, Generate validation information from information about the machine learning model. The calculated verification information is sent to a distributed database, thereby causing the distributed database to store the verification information. child It is characterized by the following.

[0013] Furthermore, in order to achieve the above objectives, the second aspect of this disclosure program teeth, On the computer, We retrieve validation information about machine learning models stored in a distributed database. Determine whether the verification information obtained from outside the distributed database is the same as the verification information regarding the machine learning model stored in the distributed database, and verify the machine learning model. Send the verification result to the distributed database, thereby storing the verification information in the distributed database. child It is characterized by the above.

Advantages of the Invention

[0014] As described above, according to the present disclosure, it is possible to prevent tampering without publicly releasing the data used for the approval of medical devices.

Brief Description of the Drawings

[0015] [Figure 1] FIG. 1 is a configuration diagram showing a schematic configuration of an information processing system, a first terminal device, and a second terminal device in an embodiment. [Figure 2] FIG. 2 is a diagram showing an example of information stored in a distributed database and a terminal device in an embodiment. [Figure 3] FIG. 3 is a flowchart showing the operation of the first terminal device in an embodiment. [Figure 4] FIG. 4 is a flowchart showing the operation of the second terminal device in an embodiment. [Figure 5] FIG. 5 is a block diagram showing an example of a computer that realizes a terminal device in an embodiment.

Modes for Carrying Out the Invention

[0016] (Embodiment) Hereinafter, an information processing system, a terminal device, an information processing method, and a program in an embodiment will be described with reference to FIGS. 1 to 5.

[0017] [Device Configuration] First, the schematic configuration of the information processing system and terminal device in the embodiment will be described using Figure 1. Figure 1 is a configuration diagram showing the schematic configuration of the information processing system, the first terminal device, and the second terminal device in the embodiment.

[0018] The information processing system 100 in the embodiment shown in Figure 1 includes a distributed database 10. The distributed database 10 stores verification information related to machine learning models. This verification information is transmitted from the first terminal device 30. Furthermore, this verification information is transmitted from the distributed database 10 to the second terminal device 40 in response to a request from the second terminal device 40.

[0019] Furthermore, as shown in Figure 1, the first terminal device 30 and the second terminal device 40 are connected to a distributed database 10 via a network 20 such as the Internet. The first terminal device 30 includes a verification information generation unit 31 and a verification information transmission unit 32. The verification information generation unit 31 generates verification information from information related to the machine learning model. Verification information Generation section 31 It functions as a means of generating verification information.

[0020] The verification information transmission unit 32 transmits the calculated verification information to the distributed database 10 for storage. The verification information transmission unit 32 functions as a verification information transmission means.

[0021] The second terminal device 40 includes a verification information acquisition unit 41, a verification unit 42, and a verification result transmission unit 43. The verification information acquisition unit 41 acquires verification information about machine learning models stored in the distributed database 10. The verification information acquisition unit 41 functions as a verification information acquisition means.

[0022] The verification unit 42 verifies the machine learning model by determining whether the verification information obtained from sources other than the distributed database is the same as the verification information about the machine learning model obtained by the verification information acquisition unit 41. The verification unit 42 functions as a verification means.

[0023] The verification result transmission unit 43 transmits the verification results from the verification unit 42 to the distributed database 10 for storage. The verification result transmission unit 43 functions as a verification result transmission means.

[0024] Thus, the first terminal device transmits only the verification information for data used in the approval of medical devices, that is, information related to machine learning, to the distributed database 10. The second terminal device then retrieves the verification information stored in the distributed database 10 and uses the retrieved verification information to verify the identity of the information related to machine learning.

[0025] In other words, the information processing system 100 in this embodiment can provide a mechanism for exchanging data between medical device manufacturers and verifiers without disclosing the data used for medical device approval and without preventing its tampering.

[0026] Next, the configuration and functions of the information processing system 100, the first terminal device 30, and the second terminal device 40 in this embodiment will be described in detail.

[0027] First, the information processing system 100 is a data management system for managing data, particularly highly confidential data, and, as described above, includes a distributed database 10. In this embodiment, the information processing system 100 is used to manage data necessary for the approval of medical devices. However, the use of the information processing system 100 is not particularly limited.

[0028] Furthermore, in this embodiment, the distributed database 10 is constructed by multiple nodes connected via a network. For example, the distributed database 10 is constructed by multiple nodes 11 using distributed ledger technology. In this case, each node 11 communicates data with other nodes 11 on a peer-to-peer basis to form a distributed ledger. Some distributed ledgers have a function to prevent data tampering by having multiple nodes hold data and verifying the data held by each node with one another. Another example of such a distributed ledger is a system using blockchain.

[0029] Each node 11 is actually connected to the network 20 in a data communication manner and communicates with other nodes 11 peer-to-peer via the network 20. Specific examples of nodes 11 include general-purpose personal computers (PCs), server devices, smartphones, and tablet devices.

[0030] Furthermore, in this embodiment, the user of the first terminal device 30 is the creator of the machine learning model, that is, the manufacturer of the medical device equipped with the machine learning model. The user of the second terminal device 40 is the person who verifies the medical device, for example, the person in charge at the PMDA or similar organization mentioned above.

[0031] Furthermore, in this embodiment, in addition to the first terminal device 30 and the second terminal device 40, a third terminal device 50 is also connected to the distributed database 10 via the network 20. The user of the third terminal device 50 is an approver who approves medical devices using the results of the verification performed by the second terminal device 40. An example of an approver is a person in charge at an administrative agency that approves medical devices.

[0032] Here, we will specifically explain the transmission and reception of information in the information processing system 100 using Figure 2. Figure 2 is a diagram showing an example of information stored in a distributed database and terminal device in this embodiment.

[0033] In the example shown in Figure 2, the information about the machine learning model includes documentation data about the machine learning model, test data input to the machine learning model, training data used to build the machine learning model, the learning program used to build the machine learning model, and environmental information.

[0034] Document data related to machine learning models includes information about medical devices that utilize the machine learning models, such as clinical trial plans. Environmental information identifies the environment in which machine learning is performed and includes, for example, hardware specifications and operating system versions.

[0035] As shown in Figure 2, the medical device manufacturer prepares document data, test data, training data, learning programs, and environmental information related to the machine learning model in the first terminal device 30.

[0036] In the first terminal device 30, the verification information generation unit 31 generates verification information from the document data, test data, training data, learning program, and environment information related to the machine learning model.

[0037] Verification information is generated from the data being verified and is used to verify or prove the identity of the original data. In this embodiment, identical data will generate identical verification information. Furthermore, if the data is different, the generated verification information will also be different with a practically sufficient probability. Therefore, it is difficult for a third party to infer the original data from the verification information, and it is also difficult for a third party to find the data necessary to generate identical verification information. A concrete example of such verification information is a hash value.

[0038] A method for verifying or proving the identity of information about a machine learning model using verification information is as follows: First, a specific terminal device generates verification information about the machine learning model in advance and saves the generated verification information. Next, another terminal device generates verification information about the machine learning model at a later point in time and determines whether it is identical to the verification information saved earlier. If they are identical, it can be determined that the information about the machine learning model at the time the verification information was saved is identical to the information about the machine learning model at the time of the determination. This makes it possible to verify or prove that the information about the machine learning model has not been tampered with since the verification information was saved.

[0039] Specifically, the verification information generation unit 31 calculates hash values ​​from information related to the machine learning model, namely, document data, test data, training data, learning program, and environment information related to the machine learning model. Hereafter, the verification information will be referred to as "hash value." The verification information transmission unit 32 then transmits each hash value, which constitutes the verification information, to the distributed database 10.

[0040] As a result, the distributed database 10 stores each hash value transmitted from the first terminal device 30. In Figure 2, the information written in white text on a black background shows the hash values ​​calculated from the original information.

[0041] Furthermore, the first terminal device 30 transmits documentation data, test data, training data, learning program, and environmental information related to the machine learning model to the second terminal device 40 in accordance with the instructions of the second terminal device 40. The second terminal device 40 stores the transmitted documentation data, test data, training data, learning program, and environmental information related to the machine learning model.

[0042] In the second terminal device 40, the verification information acquisition unit 41 acquires document data, test data, training data, learning programs, and environmental information related to the machine learning model, as well as their respective hash values, from the distributed database 10.

[0043] In the second terminal device 40, the verification unit 42 calculates hash values ​​for each of the document data, test data, training data, learning program, and environmental information related to the machine learning model that have been transmitted from the first terminal device 30. The verification unit 42 then determines whether each of the calculated hash values ​​is the same as each of the hash values ​​obtained from the distributed database 10. Subsequently, the verification result transmission unit 43 transmits the determination result for each hash value as the verification result to the distributed database 10.

[0044] In this embodiment, the distributed database 10 also stores the verification results transmitted from the second terminal device 40. That is, the distributed database 10 also stores the results of the verification performed by the second terminal device 40 on the machine learning model.

[0045] The third terminal device 50 stores the manufacturer's confidential information and the verifier's confidential information. Upon instruction from the certifying party, the third terminal device 50 accesses the distributed database 10 using the manufacturer's confidential information and the verifier's confidential information. If the distributed database 10 finds that the manufacturer's confidential information and the verifier's confidential information match the authentication information stored in advance, it grants access to the third terminal device 50 and transmits the verification results. This allows the certifying party to decide whether to certify the medical device based on the verification results.

[0046] In this embodiment, unlike the example in Figure 2, the first terminal device 30 may transmit the hash values ​​of the document data, test data, training data, learning program, and environment information related to the machine learning model to the second terminal device 40. In this case, the verification unit 42 in the second terminal device determines whether each hash value transmitted from the first terminal device 30 is the same as each hash value obtained from the distributed database 10.

[0047] Furthermore, in this embodiment, the first terminal device 30 can also transmit the document data itself to the distributed database 10 without calculating a hash value for the document data related to the machine learning model.

[0048] [Device operation] Next, the operation of the information processing system and terminal device in the embodiment will be described using Figures 3 and 4. In the following description, Figures 1 and 2 will be referred to as appropriate. In the embodiment, the information processing method is carried out by operating the information processing system 100, the first terminal device 30, and the second terminal device 40. Therefore, the description of the information processing method in the embodiment will be replaced by the following description of the operation of the information processing system 100, the first terminal device 30, and the second terminal device 40.

[0049] First, the operation of the first terminal device 30 will be explained using Figure 3. Figure 3 is a flowchart showing the operation of the first terminal device in the embodiment.

[0050] As shown in Figure 3, first, in the first terminal device 30, the verification information generation unit 31 calculates hash values ​​as verification information from information related to the machine learning model, specifically from document data, test data, training data, learning program, and environment information (Step A1).

[0051] Next, in the first terminal device 30, the verification information transmission unit 32 transmits each hash value calculated in step A1 to the distributed database 10 (step A2).

[0052] After step A2 is executed, the distributed database 10 stores each hash value sent from the first terminal device 30 (see Figure 2).

[0053] Next, the operation of the second terminal device 40 will be explained using Figure 4. Figure 4 is a flowchart showing the operation of the second terminal device in the embodiment.

[0054] As shown in Figure 4, first, in the second terminal device 40, the verification information acquisition unit 41 acquires the hash values ​​of document data, test data, training data, learning program, and environment information related to the machine learning model from the distributed database 10 (step B1).

[0055] Next, in the second terminal device 40, the verification unit 42 instructs the first terminal device 30 to transmit information related to the machine learning model, namely document data, test data, training data, learning program, and environment information (step B2).

[0056] Next, the verification unit 42 calculates hash values ​​from the document data, test data, training data, learning program, and environment information related to the machine learning model that have been transmitted from the first terminal device 30 after the execution of step B2 (step B3).

[0057] Next, the verification unit 42 determines whether each hash value calculated in step B3 is the same as each hash value obtained in step B1 (step B4).

[0058] Subsequently, in the second terminal device 40, the verification result transmission unit 43 transmits the determination result for each hash value in step B4 as a verification result to the distributed database 10 for storage (step B5).

[0059] After step B5 is completed, the certifier obtains the verification results via the third terminal device 50. Based on the obtained verification results, the certifier then makes a decision regarding the approval of the medical device.

[0060] Although the operation of the second terminal device was explained using Figure 4, the processes described in Figure 4 do not necessarily have to be performed during the verification process of medical devices. For example, if the verifier does not need information about the machine learning model, they do not need to have the manufacturer provide that information. Also, the verifier, who is the user of the second terminal device 40, may instruct in step B2 to transmit only a portion of the information about the machine learning model. In this way, information about the machine learning model is not disclosed when not needed, and when necessary, it is handled in a way that allows for verification of whether or not it has been tampered with.

[0061] As described above, according to the embodiment, it becomes possible to exchange data between manufacturers and verifiers, and further between verifiers and certifiers, without disclosing information used for the approval of medical devices, i.e., information related to machine learning models, and while preventing tampering.

[0062] [program] The first program in this embodiment can be any program that causes a computer to execute steps A1 and A2 shown in Figure 3. By installing and running this program on a computer, the first terminal device 30 in this embodiment can be realized. In this case, the computer's processor functions as a verification information generation unit 31 and a verification information transmission unit 32, and performs processing. Examples of computers include general-purpose PCs, smartphones, and tablet terminal devices.

[0063] The second program in the embodiment can be any program that causes a computer to execute steps B1 to B5 shown in Figure 4. By installing and running this program on a computer, the second terminal device 40 in the embodiment can be realized. In this case, the computer's processor functions as a verification information acquisition unit 41, a verification unit 42, and a verification result transmission unit 43, and performs processing. Examples of computers include general-purpose PCs, smartphones, and tablet terminal devices.

[0064] [Physical configuration] Here, a computer that implements the first terminal device 30 and the second terminal device 40 by executing the program in the embodiment will be described with reference to Figure 5. Figure 5 is a block diagram showing an example of a computer that implements the terminal devices in the embodiment.

[0065] As shown in Figure 5, the computer 110 comprises a CPU (Central Processing Unit) 111, main memory 112, storage device 113, input interface 114, display controller 115, data reader / writer 116, and communication interface 117. Each of these components is connected to the others via a bus 121, enabling data communication.

[0066] Furthermore, the computer 110 may include a GPU (Graphics Processing Unit) or an FPGA (Field-Programmable Gate Array) in addition to, or instead of, the CPU 111. In this embodiment, the GPU or FPGA can execute the program in the embodiment.

[0067] The CPU 111 loads the program in the embodiment, which consists of a set of codes stored in the storage device 113, into the main memory 112, and performs various calculations by executing each code in a predetermined order. The main memory 112 is typically a volatile storage device such as DRAM (Dynamic Random Access Memory).

[0068] Furthermore, the program in this embodiment is provided stored on a computer-readable recording medium 120. The program in this embodiment may also be distributed over the internet via a communication interface 117.

[0069] Specific examples of the storage device 113 include hard disk drives and semiconductor storage devices such as flash memory. The input interface 114 mediates data transmission between the CPU 111 and input devices 118 such as a keyboard and mouse. The display controller 115 is connected to the display device 119 and controls the display on the display device 119.

[0070] The data reader / writer 116 mediates data transmission between the CPU 111 and the recording medium 120, reads programs from the recording medium 120, and writes processing results from the computer 110 to the recording medium 120. The communication interface 117 mediates data transmission between the CPU 111 and other computers.

[0071] Furthermore, specific examples of the recording medium 120 include general-purpose semiconductor memory devices such as CF (Compact Flash®) and SD (Secure Digital), magnetic recording media such as Flexible Disks, or optical recording media such as CD-ROMs (Compact Disk Read Only Memory).

[0072] Furthermore, the first terminal device 30 and the second terminal device 40 in this embodiment can be implemented not by a computer with a program installed, but by using hardware corresponding to each part, such as electronic circuits. Moreover, the first terminal device 30 and the second terminal device 40 may each be partially implemented by a program and the remaining part by hardware. In this embodiment, the computer is not limited to the computer shown in Figure 5.

[0073] Some or all of the embodiments described above can be expressed by (Appendix 1) to (Appendix 20) described below, but are not limited to the following descriptions.

[0074] (Note 1) Equipped with a distributed database, The aforementioned distributed database stores validation information regarding machine learning models, The verification information is transmitted from the first terminal device and, in response to a request from the second terminal device, is transmitted from the distributed database to the second terminal device. Information processing system.

[0075] (Note 2) The information relating to the machine learning model includes test data input to the machine learning model, training data used to construct the machine learning model, and the learning program used to construct the machine learning model. The information processing system described in Appendix 1.

[0076] (Note 3) The information relating to the machine learning model further includes environmental information that identifies the environment during machine learning. The information processing system described in Appendix 2.

[0077] (Note 4) The distributed database further stores the results of validation performed on the machine learning model, The above result is transmitted from the second terminal device. The information processing system described in Appendix 1 or 2.

[0078] (Note 5) A validation information generation unit that generates validation information from information about the machine learning model, A verification information transmission unit transmits and stores the calculated verification information in a distributed database. A terminal device equipped with these features.

[0079] (Note 6) The information relating to the machine learning model includes test data input to the machine learning model, training data used to construct the machine learning model, and the learning program used to construct the machine learning model. The terminal device described in Appendix 5.

[0080] (Note 7) The information relating to the machine learning model further includes environmental information that identifies the environment during machine learning. The terminal device described in Appendix 6.

[0081] (Note 8) A validation information acquisition unit that retrieves validation information about machine learning models stored in a distributed database, A verification unit that determines whether the verification information obtained from sources other than the distributed database and the verification information obtained by the verification information acquisition unit regarding the machine learning model are identical, and verifies the machine learning model. A verification result transmission unit transmits and stores the verification results from the verification unit to the distributed database. A terminal device equipped with these features.

[0082] (Note 9) The information relating to the machine learning model includes test data input to the machine learning model, training data used to construct the machine learning model, and the learning program used to construct the machine learning model. The terminal device described in Appendix 8.

[0083] (Note 10) The information relating to the machine learning model further includes environmental information that identifies the environment during machine learning. The terminal device described in Appendix 9.

[0084] (Note 11) The first terminal device generates validation information from information about the machine learning model and transmits the calculated validation information to a distributed database. The distributed database stores the transmitted verification information and transmits the verification information to the second terminal device in response to a request from the second terminal device. The second terminal device acquires the verification information stored in the distributed database, determines whether the verification information obtained from sources other than the distributed database is the same as the verification information stored in the distributed database, and verifies the machine learning model. Information processing methods.

[0085] (Note 12) The information relating to the machine learning model includes test data input to the machine learning model, training data used to construct the machine learning model, and the learning program used to construct the machine learning model. The information processing method described in Appendix 11.

[0086] (Note 13) The information relating to the machine learning model further includes environmental information that identifies the environment during machine learning. The information processing method described in Appendix 12.

[0087] (Note 14) The second terminal device further transmits the verification results to the distributed database and has them stored in the distributed database. The information processing method described in Appendix 11 or 12.

[0088] (Note 15) On the computer, Generate validation information from information about the machine learning model. The calculated verification information is sent to a distributed database, thereby causing the distributed database to store the verification information. P Logra Hmm.

[0089] (Note 16) The information relating to the machine learning model includes test data input to the machine learning model, training data used to construct the machine learning model, and the learning program used to construct the machine learning model. As described in Appendix 15 The program .

[0090] (Note 17) The information relating to the machine learning model further includes environmental information that identifies the environment during machine learning. As described in Appendix 16 program .

[0091] (Note 18) On the computer, We retrieve validation information about machine learning models stored in a distributed database. The system determines whether the verification information obtained from sources other than the aforementioned distributed database is identical to the verification information regarding the machine learning model stored in the aforementioned distributed database, and then verifies the machine learning model. The verification results are sent to the distributed database, thereby causing the verification information to be stored in the distributed database. P Logra Hmm.

[0092] (Note 19) The information relating to the machine learning model includes test data input to the machine learning model, training data used to construct the machine learning model, and the learning program used to construct the machine learning model. As described in Appendix 18 program .

[0093] (Note 20) The information relating to the machine learning model further includes environmental information that identifies the environment during machine learning. As described in Appendix 19 program .

[0094] Although the present invention has been described above with reference to embodiments, the present invention is not limited to the above embodiments. Various modifications to the structure and details of the present invention can be made, as can be understood by those skilled in the art within the scope of the present invention. [Industrial applicability]

[0095] As described above, this disclosure makes it possible to prevent tampering with data used in the approval of medical devices without disclosing it. This disclosure is useful in various fields where it is necessary to prevent tampering with data without disclosing it. [Explanation of symbols]

[0096] 10. Distributed databases 20 Networks 30 First terminal device 31 Verification Information Generation Unit 32 Verification Information Transmission Unit 40 Second terminal device 41 Verification Information Acquisition Unit 42 Verification Department 43 Verification Result Transmission Unit 50 Third terminal device 100 Information Processing Systems 110 Computer 111 CPU 112 Main Memory 113 Storage device 114 Input Interface 115 Display Controller 116 Data Readers / Writers 117 Communication Interface 118 Input devices 119 Display device 120 recording media 121 Bus

Claims

1. Equipped with a distributed database, The aforementioned distributed database stores validation information regarding machine learning models, The verification information is transmitted from the first terminal device and, in response to a request from the second terminal device, is transmitted from the distributed database to the second terminal device. The second terminal device acquires the verification information stored in the distributed database, determines whether the verification information obtained from sources other than the distributed database is the same as the verification information stored in the distributed database, and verifies the machine learning model. The information relating to the machine learning model includes test data input to the machine learning model, training data used to construct the machine learning model, a learning program used to construct the machine learning model, and environmental information that identifies the environment during machine learning. The verification information is a hash value calculated from the test data, the training data, the learning program, and the environmental information, respectively. Information processing system.

2. The distributed database further stores the results of validation performed on the machine learning model, The above result is transmitted from the second terminal device. The information processing system according to claim 1.

3. A validation information acquisition unit that retrieves validation information about machine learning models stored in a distributed database, A verification unit that determines whether the verification information obtained from sources other than the distributed database and the verification information obtained by the verification information acquisition unit regarding the machine learning model are identical, and verifies the machine learning model. A verification result transmission unit transmits and stores the verification results from the verification unit to the distributed database. Equipped with, The information relating to the machine learning model includes test data input to the machine learning model, training data used to construct the machine learning model, a learning program used to construct the machine learning model, and environmental information that identifies the environment during machine learning. The verification information is a hash value calculated from the test data, the training data, the learning program, and the environmental information, respectively. Terminal device.

4. The first terminal device generates validation information from information about the machine learning model and transmits the calculated validation information to a distributed database. The distributed database stores the transmitted verification information and transmits the verification information to the second terminal device in response to a request from the second terminal device. The second terminal device acquires the verification information stored in the distributed database, determines whether the verification information obtained from sources other than the distributed database is the same as the verification information stored in the distributed database, and verifies the machine learning model. The information relating to the machine learning model includes test data input to the machine learning model, training data used to construct the machine learning model, a learning program used to construct the machine learning model, and environmental information that identifies the environment during machine learning. The verification information is a hash value calculated from the test data, the training data, the learning program, and the environmental information, respectively. Information processing methods.

5. The second terminal device further transmits the verification results to the distributed database and stores them in the distributed database. The information processing method according to claim 4.

6. On the computer, We retrieve validation information about machine learning models stored in a distributed database. The system determines whether the verification information obtained from sources other than the aforementioned distributed database is identical to the verification information regarding the machine learning model stored in the aforementioned distributed database, and then verifies the machine learning model. The verification results are sent to the distributed database, thereby causing the verification results to be stored in the distributed database. The information relating to the machine learning model includes test data input to the machine learning model, training data used to construct the machine learning model, a learning program used to construct the machine learning model, and environmental information that identifies the environment during machine learning. The verification information is a hash value calculated from the test data, the training data, the learning program, and the environmental information, respectively. program.

Images