Information processing system

Abstract

To provide an information processing system capable of securing confidentiality of an inference model while protecting privacy of medical data.SOLUTION: An information processing system 1 includes a first information processing apparatus and a second information processing apparatus configured to communicate over a network. The first information processing apparatus includes a first partial model configured to include a first input layer, a first intermediate layer which is at least one of intermediate layers in a group of intermediate layers, and a first output layer. The second information processing apparatus includes a second partial model configured to include a second intermediate layer including an intermediate layer different from the intermediate layer constituting the first intermediate layer in the group of intermediate layers, and a second output layer. The second information processing apparatus includes: an inference target data acquisition unit configured to acquire inference target data; and an inference unit configured to perform predetermined inference processing on the inference target data by using the first partial model and the second partial model.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an information processing system that performs inference processing or learning processing using an inference model. [Background technology]

[0002] Systems are known that apply machine learning techniques to medical data, such as medical images acquired by medical imaging devices (modalities) and medical information acquired from medical information systems, to make inferences about specific diseases (disease detection, benign / malignant distinction, prognosis prediction, risk prediction, etc.).

[0003] Non-patent document 1 discloses a technology in which, when an image is input, a first local information processing device compresses the image and the compressed result is transferred to a second information processing device in a remote cloud service to perform disease segmentation.

[0004] Non-Patent Document 2 discloses a technology in which a model consisting of an encoder and decoder is trained using a medical image as input, and feature extraction is performed using an inference model (encoder section). [Prior art documents] [Non-patent literature]

[0005] [Non-Patent Document 1] Zihao Liu et al., “Orchestrating Medical Image Compression and Remote Segmentation Networks”, International Conference on Medical Image Computing and Computer-Assisted Intervention-MICCAI2020. [Non-Patent Document 2] Behzad Bozorgtabar et al., “SALAD: Self-Supervised Aggregation Learning for Anomaly Detection on X-Rays”, International Conference on Medical Image Computing and Computer-Assisted Intervention-MICCAI2020. Summary of the Invention [Problem to be solved by the invention]

[0006] In the technology described in Non-Patent Document 1, when performing segmentation of a medical image, the medical image, which is the data to be inferred, is transmitted to a cloud service. In this case, since medical data such as medical images are transmitted from within the hospital to an external information processing device, it is difficult to protect the privacy of the medical data.

[0007] On the other hand, as in the technology described in Non-Patent Document 2, there are cases where an inference model is built in an information processing device managed by a user who performs inference. In such cases, it is difficult to prevent the user from copying and distributing the inference model to a third party, or from improperly using the inference model, such as by modifying the inference model, and therefore the confidentiality of the inference model cannot be guaranteed.

[0008] The present invention aims to provide an information processing system that can protect the privacy of medical data while ensuring the confidentiality of inference models. [Means for solving the problem]

[0009] In order to solve the above problem, an information processing system according to one aspect of the present invention comprises a first information processing device and a second information processing device capable of communicating with the first information processing device via a network, and performs a predetermined inference processing using an inference model based on a neural network including a first input layer, an intermediate layer group, a first output layer, and a second output layer, the first output layer and the second output layer being provided in different information processing devices, the first information processing device has a first partial model including the first input layer and a first intermediate layer which is at least a part of the intermediate layers of the intermediate layer group, and the first output layer, the second information processing device has a second partial model including a second intermediate layer including an intermediate layer of the intermediate layer group different from the intermediate layer constituting the first intermediate layer, and the second output layer, and the information processing system has an inference target data acquisition unit that acquires inference target data, and an inference unit that performs a predetermined inference processing on the inference target data using the first partial model and the second partial model. Effect of the Invention

[0010] According to the present invention, it is possible to protect the privacy of medical data while ensuring the confidentiality of inference models. [Brief description of the drawings]

[0011] [Figure 1] FIG. 1 is a diagram showing the configuration of an information processing system according to a first embodiment and a third embodiment. [Diagram 2] FIG. 2 is a diagram showing a hardware configuration of a first information processing apparatus according to the first embodiment. [Diagram 3] Schematic diagram of an inference model relating to the first embodiment. [Figure 4] FIG. 4 is a flowchart showing an inference process of the information processing system according to the first embodiment. [Diagram 5] FIG. 1 is a diagram showing the configuration of an information processing system according to a second embodiment and a fourth embodiment. [Figure 6] Schematic diagram of an inference model relating to the second embodiment. [Figure 7]A flow chart showing the learning process of an inference model in the second embodiment. [Figure 8] A configuration diagram of an inference model for the third and fourth embodiments. [Figure 9] FIG. 11 is a flowchart showing an inference process of an information processing system according to a third embodiment. [Figure 10] FIG. 13 is a flowchart showing a learning process of an information processing system according to a fourth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] The present invention can be preferably applied to medical data such as raw data (signal data) acquired by a modality and diagnostic data generated from the raw data by image reconstruction. Modalities include, for example, X-ray CT scanners, MRI scanners, SPECT scanners, PET scanners, and electrocardiographs. The data to be inferred and the teacher data may be not only medical data, but also information related to the patient's privacy, such as age, gender, and disease information. The target data is not limited to medical data. For example, the target may be image data depicting a person, text data based on document data, audio data based on audio, or any other data that can be inferred using a neural network.

[0013] Hereinafter, the inference process by the inference model in the information processing system of the present invention will be described in the first and third embodiments. The first and third embodiments are different in the inference model that performs inference and the flow of the inference process. In addition, in the second and fourth embodiments, the learning process of the inference model in the information processing system of the present invention will be described. Note that, as will be described later, the inference model used in the inference process is not limited to the inference model generated through at least one of the learning processes of the second and fourth embodiments of the present invention. The inference model used in the inference process is a trained inference model trained based on machine learning or deep learning by a known method or a learning process described in the present invention. Here, the trained inference model may be a model that has been trained based on machine learning or deep learning by a learning process that satisfies a predetermined condition, and the trained inference model may be used as a target for additional learning, transfer learning, fine tuning, etc. Therefore, a learning process by a learning process described later may be performed as additional learning of a trained inference model trained by a known method, or the learning process may be performed in the reverse order.

[0014] Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings.

[0015] [First embodiment] Here, an information processing system 1 of the present invention will be described with reference to Fig. 1. The information processing system 1 of the present invention includes a first information processing device 2 and a second information processing device 3 capable of communicating with the first information processing device 2 via a network.

[0016] 2 shows an example of a specific configuration of the first information processing device 2. In this example, the first information processing device 2 has a CPU 200, a GPU 201, a RAM 202, a ROM 203, and a storage device 204, which are connected via a system bus 205. In addition, a display device 206 and an input device 207 such as a mouse and a keyboard are connected to the first information processing device 2. The second information processing device 3 may be configured in a similar manner, or may be configured from a part of the configuration of the first information processing device 2.

[0017] The functional configuration of the information processing system 1 and the inference model in the present invention will be described below with reference to FIG. 3. The information processing system 1 includes a first information processing device 2 and a second information processing device 3 that can communicate with the first information processing device 2 via a network. Here, for example, the first information processing device 2 is an information processing device managed by a user of the inference model, and the second information processing device 3 is an information processing device managed by a provider of the inference model. The first information processing device 2 and the second information processing device 3 each have a partial model that is a part of the inference model that performs inference processing on medical data to be inferred and outputs the execution result. The inference model is a neural network including a first input layer, a middle layer group, a first output layer, and a second output layer, and the first output layer is provided as a configuration of the partial model of the first information processing device 2. The second output layer is provided as a configuration of the partial model of the second information processing device 3. In the inference processing step, the inference model is a trained inference model that has been trained through a predetermined learning process. A parameter for outputting an inference result is determined by the learning process, and a model in which the parameter and the network model are paired is defined as the inference model.

[0018] The first information processing device 2 has a first partial model including a first input layer, a first intermediate layer which is at least a part of the intermediate layers in the intermediate layer group, and a first output layer among the above-mentioned inference models. For example, as the first intermediate layer, the first information processing device 2 has N1 intermediate layers from the first intermediate layer to the N1 intermediate layer as shown in FIG. 3.

[0019] The second information processing device 3 also has a second partial model including a second intermediate layer including an intermediate layer different from the intermediate layer constituting the first intermediate layer among the intermediate layer group, and a second output layer. For example, as the second intermediate layer, as shown in FIG. 3, it has N2 intermediate layers from the N1+1th intermediate layer to the N1+N2th intermediate layer. Here, the first information processing device 2 and the second information processing device 3 may each have an intermediate layer with common parameters. For example, in addition to the above, the first intermediate layer and the second intermediate layer may have N3 common intermediate layers from the N1+N2+1th intermediate layer to the N1+N2+N3th intermediate layer not shown in FIG. 3. The configuration of each information processing device will be described below.

[0020] The first information processing device 2 is an information processing device that can be operated by a user of the inference model who has the authority to manage the medical data to be inferred, for example, a medical professional. On the other hand, the second information processing device 3 is an information processing device owned by a provider of the model who has the authority to manage the inference model used for inference. The second information processing device 3 exists in a server outside the first information processing device 2, and is configured to be able to communicate via a network 4.

[0021] The first information processing device 2 has a storage unit 20 that stores model information of the first partial model, inference target data, etc. The storage unit 20 may be configured from a device external to the first information processing device 2. The first information processing device is also configured to include an inference target data acquisition unit 21 that acquires medical data of the inference target, and a first inference unit 22 that performs a first inference process using the first partial model, among inference units that perform inference processing on the medical data of the inference target.

[0022] The second information processing device 3 is configured to include a second inference unit 31, among the inference units that perform inference processing, which exists in a server external to the first information processing device 2 and receives the result of the first inference processing from the first information processing device 2, and performs a second inference processing using a second partial model.

[0023] The information processing system 1 according to the present invention performs inference processing in which an inference model based on a neural network including a first input layer, a group of intermediate layers, a first output layer, and a second output layer is divided into a first partial model of a first information processing device 2 and a second partial model of a second information processing device 3, thereby ensuring the confidentiality of the inference model while protecting the privacy of medical data. Specifically, a user of the inference model, who is the administrator of the data to be inferred, does not need to transmit the medical data to be inferred to an external information processing device, and can obtain an inference result while protecting the privacy of the medical data.

[0024] In addition, the provider of the inference model can ensure the confidentiality of the inference model by installing only a portion of the inference model in the first information processing device 2.

[0025] Here, the inference process performed by the information processing system 1 in this embodiment will be described with reference to FIG.

[0026] In step S40, the inference target data acquisition unit 121 in the first information processing device 2 acquires medical data of the inference target, transmits the acquired medical data of the inference target to the first inference unit 22 that constitutes the inference unit, and proceeds to the next step.

[0027] In step S41, the inference unit performs a predetermined inference process using the first partial model and the second partial model. Specifically, the medical data to be inferred is input to the first input layer constituting the first inference unit 22, inference processes are performed on the first hidden layer to the N1th hidden layer, and intermediate information is transmitted to the second inference unit 31. The second inference unit 31 receives the intermediate information transmitted from the first inference unit 22, and performs inference processes on the N1+1th hidden layer to the N1+N2th hidden layer. Then, the inference result on the N1+N2th hidden layer is transmitted to the second output layer constituting the second partial model and the first output layer constituting the first partial model, and the process proceeds to the next step.

[0028] In this embodiment, N1+N2 is the final layer that constitutes the intermediate layer group, but N1+N2 does not have to be the final intermediate layer, and N1+N2+1 to the final intermediate layer may be configured as a common intermediate layer in each partial model. At this time, the inference result by N1+N2 intermediate layer is sent to N1+N2+1 layer that constitutes each partial model. Then, the process from each N1+N2+1 layer to the final intermediate layer is executed, and the inference result by the final intermediate layer is sent to each output layer.

[0029] In step S42, the first output layer of the first partial model and the second output layer of the second partial model each output an inference result for a predetermined inference process using intermediate information from the N+2th intermediate layer as an input. The inference result is also output to the second information processing device 3, which is the distributor of the inference model, to prevent unintended inference. Specifically, when an image unintended by the provider of the inference model, which is the distributor, is input by a user who is the recipient of the inference model, for example, when inference is performed multiple times with a small amount of noise information added, the inference process can be detected from the likelihood distribution of the output by the second output layer and the difference in the results of multiple inference processes by the second output layer. When the second information processing device 3 detects an unintended inference process, it stops the use of the inference model. Specifically, it does not permit the input of information to the intermediate layer provided only in the second information processing device 3. With this configuration, it is possible to ensure the confidentiality of information in the intermediate layer constituting the second partial model of the second information processing device 3.

[0030] By implementing the above-mentioned inference model and inference processing process, the information processing system 1 described in the present invention can protect the privacy of the medical data that is the subject of inference while ensuring the confidentiality of the inference model that performs inference on the medical data that is the subject of inference.

[0031] [Second embodiment] In the first embodiment, an inference process using an inference model based on a neural network including a first input layer, a group of intermediate layers, a first output layer, and a second output layer has been described. In the present embodiment, a learning process using the inference model will be described.

[0032] The functional configuration of the information processing system 1 in this embodiment will be described with reference to Fig. 5. The learning process may be performed before the inference of the inference model in the first embodiment, or the inference model may be additionally learned through the learning process after the inference.

[0033] In this embodiment, the information processing system 1 has a storage unit 40 that stores teacher data and information on an inference model. The information processing system 1 further includes a teacher data acquisition unit 41 that acquires teacher data from the storage unit 40, and a first learning unit 42 that learns a first partial model based on the acquired teacher data. Here, the learning unit in the information processing system 1 includes the first learning unit 42 that learns the first partial model, and a second learning unit 51 that learns the second partial model.

[0034] The learning unit is a learning unit that inputs learning data constituting the teacher data to a first input layer of the inference model and learns the intermediate layer group and the first output layer, and learns the intermediate layer group and the first output layer by an error backpropagation method using loss information calculated using the correct answer data constituting the teacher data and the first output layer, and learns the inference model by using the parameters related to the first output layer obtained by the learning as parameters related to the second output layer.

[0035] By being configured in this manner, the information processing system 1 can perform learning processing of an inference model while ensuring the confidentiality of medical data such as training data.

[0036] The learning process of the inference model in this embodiment will be explained below with reference to FIGS.

[0037] In step S70, the teacher data acquisition unit 41 acquires teacher data in which learning data and correct answer data are paired from the storage unit 40. The teacher data acquisition unit 41 transmits information about the learning data to the first learning unit 42, and proceeds to the next step.

[0038] In step S71, the first learning unit 42 inputs learning data to the first input layer, propagates the data forward through the intermediate layer group, which is the first intermediate layer, the N1th intermediate layer, the N1+12th intermediate layer, and the N1Lth intermediate layer, calculates loss information using the output obtained through the first output layer and the correct data, and proceeds to the next step.

[0039] In step S72, the first learning unit 42 learns the first output layer and the hidden layer group by the backpropagation method using the calculated loss information, and then transmits parameters related to the first output layer to the second learning unit 51 and proceeds to the next step. Note that the learning process up to this step may be repeated for a predetermined number of learning data or epochs before proceeding to the next step, or the process of the next step may be executed for each epoch or for each predetermined learning process.

[0040] In step S73, the second learning unit 52 determines the parameters for the second output layer by reusing or copying the parameters for the first output layer as the parameters for the second output layer, and ends the learning process. Note that the parameters for the second output layer may be determined on the second information processing device 4 side using the loss information by the first learning unit 42, instead of being a copy of the parameters for the first output layer.

[0041] With this configuration, the correct answer data and the learning data that constitute the teacher data do not need to be transmitted from the device on the user side of the inference model to the outside, so that the confidentiality of medical data such as teacher data can be guaranteed. Furthermore, since a part of the inference model is kept secret in the second information processing device 3, which is the information processing device on the provider side of the inference model, the confidentiality of the inference model can also be guaranteed.

[0042] [Third embodiment] In this embodiment, an inference process using an inference model having a network configuration different from that of the first embodiment will be described. Note that the description of the same configuration as that of the first embodiment will be omitted as appropriate. Regarding the functional configuration, in addition to that of FIG. 1, the second information processing device 3 may further include an acquisition unit for inference target data for verification.

[0043] Here, the inference model in this embodiment will be described with reference to FIG. 8. The inference model in this embodiment is an inference model based on a neural network including a first input layer, a second input layer, a first intermediate layer, an N1 intermediate layer, an N1+1 intermediate layer, an N1+N2 intermediate layer, an N1+N2+1 intermediate layer, an N1+N2+N3 intermediate layer, and a first output layer and a second output layer. The information processing system 1 arranges a first partial model having a part of the configuration of the inference model in a first information processing device 2, and arranges a second partial model having a part of the configuration of the inference model in a second information processing device. The first partial model is also arranged to include a first input layer, a second input layer, and a first intermediate layer and a first output layer, which are at least a part of the intermediate layers in the intermediate layer group. For example, as the first intermediate layer, as shown in FIG. 8, there are N1 intermediate layers from the first intermediate layer to the N1 intermediate layer, and N3 intermediate layers from the N1+N2+1 intermediate layer to the N1+N2+N3 intermediate layer. The second partial model includes a second input layer and a second intermediate layer including an intermediate layer group different from the first intermediate layer among the intermediate layer group, and a second output layer. For example, as shown in FIG. 8, the second intermediate layer has N1+N2 intermediate layers from the first intermediate layer to the N1+N2 intermediate layer. In this case, the N1+1 intermediate layer to the N1+N2 intermediate layer are an intermediate layer group different from the first intermediate layer. When a user performs an inference process using an inference model, the inference process is performed using the first input layer, intermediate layer group, and first output layer provided in the first information processing device 2. On the other hand, when a provider of the model performs an inference process using inference target data for verification, the inference process is performed using the second input layer, intermediate layer group, and second output layer.

[0044] It should be noted that the intermediate layers constituting the intermediate layer group are not limited to this pattern.

[0045] For example, the first partial model may not have N3 intermediate layers from the N1+N2+1th to the N1+N2+N3th intermediate layers. Also, the second partial model in the server-side second information processing device 3 may have the first, N1th, N1+1th, N1+N2+1th, and N1+N2+N3th intermediate layers that are an intermediate layer group constituting a neural network, and the first partial model on the user side of the inference model may be a partial model composed of some of the intermediate layers of the intermediate layer group. With this configuration, the server-side inference model can be verified by the server-side information processing device.

[0046] Furthermore, the number of intermediate layers that make up the intermediate layer group and the configurations of each information processing device can be combined as appropriate, but by providing the first to N1th intermediate layer groups in the first partial model that makes up the first information processing device 2, which is an information processing device on the user side of the inference model, inference and learning become possible without transmitting medical data to the information processing device on the provider side of the inference model.

[0047] Furthermore, the number of intermediate layers constituting the intermediate layer group may be any number, and data may be exchanged between the information processing devices via the intermediate layers a plurality of times.

[0048] By being configured in this manner, the information processing system 1 can not only ensure the confidentiality of the inference target data and the inference model, but also ensure the confidentiality of the inference target data for verification by the provider of the inference model.

[0049] The inference flow by the information processing system 1 in this embodiment will be described with reference to FIG.

[0050] In step S900, the inference target data acquisition unit 21 acquires the medical data of the inference target from the storage unit 20, and proceeds to the next step.

[0051] In step S901, the inference unit determines a corresponding path depending on whether the medical data to be inferred is acquired or input from the first input layer of the first partial model constituting the first information processing device 2 or the second input layer of the second partial model constituting the second information processing device 3. Specifically, when the medical data to be inferred is input to the first input layer, an inference process is performed using a first path that performs an inference process using an intermediate layer group and a first output layer. In the example of FIG. 8, the intermediate layer group that configures the first path is the first intermediate layer to the N1 intermediate layer of the first partial model, the N1+1 intermediate layer to the N1+N2 intermediate layer of the second partial model, and the N1+N2+1 intermediate layer to the N1+N2+N3 intermediate layer of the first partial model. In addition, when medical data to be inferred for verification is input to the second input layer, an inference process is performed using a second path that performs an inference process using an intermediate layer group and a second output layer. 8, the intermediate layer group constituting the second path is the first intermediate layer to the N1+N2 intermediate layer of the second partial model, and the N1+N2+1 intermediate layer to the N1+N2+N3 intermediate layer of the first partial model. That is, when medical data to be inferred is input to the first input layer, the inference unit advances the process to step S902, and when medical data to be inferred is input to the second input layer, the inference unit advances the process to step S903.

[0052] In step S902, the inference unit performs inference processing using the first route, using the first intermediate layer and the second intermediate layer, which are intermediate layer groups constituting the inference model, and the first output layer. With this configuration, the user of the inference model does not need to transmit the medical data to be inferred and the inference results to the second information processing device 3, which is the information processing device on the provider side of the inference model, and the provider of the inference model can ensure the confidentiality of the inference model by concealing part of the intermediate layer. After performing the inference processing, the information processing system 1 ends this flow.

[0053] In step S903, the inference unit performs inference processing using the second route, using the first and second intermediate layers, which are intermediate layer groups constituting the inference model, and the second output layer. With this configuration, the provider of the inference model can ensure the confidentiality of the inference model by concealing a part of the intermediate layer of the inference model, and can also perform inference on the inference target data for verification held by the provider, without transmitting the data to the first information processing device 2, which is the information processing device on the user side of the inference model, even when the inference model is updated by the learning process of the second embodiment or the fourth embodiment described later. With this configuration, the provider of the inference model can also exhibit the effect of preventing unintended deterioration in the accuracy of the inference model.

[0054] (Variation 1) In the above embodiment, a configuration has been described in which the inference unit outputs an inference result using either the first output layer or the second output layer.

[0055] In this modified example, the inference unit uses both the first output layer and the second output layer to output the inference result. Note that the inference process in the third embodiment is performed on the inference target data for verification. The path of the inference process is not limited to this example.

[0056] In this modified example, the tasks of the first output layer and the second output layer may be different. Specifically, the first output layer constituting the first partial model in the first information processing device 2, which is an information processing device on the user side, performs segmentation processing on image data, and the second output layer constituting the second partial model in the second information processing device 3, which is an information processing device on the provider side of the inference model, performs classification processing on image data. With this configuration, the provider of the inference model can detect unintended use of the inference model by multiple inference processes, and the user of the inference model can obtain the inference result of a highly confidential task such as segmentation only with the first information processing device 2, so that the confidentiality of the medical data can be guaranteed. Note that, as the inference model used in such a case, for example, a multi-task inference model such as Mask R-CNN, which is a known technology, can be applied. Of the multi-task inference model, some highly confidential tasks are provided in the information processing device on the user side of the inference model, and some tasks are provided in the information processing device on the provider side of the inference model, so that in addition to the confidentiality of the inference model and the inference target data, inference processing by an unintended inference model can be prevented.

[0057] [Fourth embodiment] In this embodiment, a learning process using an inference model having a network configuration different from that of the second embodiment will be described. Note that the description of the same configuration as that of the second embodiment will be omitted as appropriate. Regarding the functional configuration, in addition to that of FIG. 5, the second information processing device may have a teacher data acquisition unit and an inference target data acquisition unit for verification.

[0058] Here, the inference model in this embodiment will be described with reference to Fig. 8. As in the third embodiment, the inference model in this embodiment is an inference model based on a neural network including a first input layer, a second input layer, an Nth intermediate layer, an N+1th intermediate layer, an N+2th intermediate layer, and an N+3th intermediate layer, which are an intermediate layer group, and further a first output layer and a second output layer.

[0059] The information processing system 1 places a first partial model having a part of the configuration of the inference model in a first information processing device 2, and places a second partial model having a part of the configuration of the inference model in a second information processing device. The first partial model is configured to include a first input layer, a second input layer, a first intermediate layer which is at least a part of the intermediate layer group, and a first output layer. The second partial model is configured to include a second input layer, a second intermediate layer which includes an intermediate layer group different from the first intermediate layer in the intermediate layer group, and a second output layer.

[0060] When a user learns an inference model, a learning process is carried out using the first input layer, intermediate layer group, and first output layer of the inference model, which are provided in the first information processing device 2, which is the information processing device on the user's side, and the second output layer and the first intermediate layer in the second information processing device 3 complete the learning process by reusing the parameters related to the first output layer determined by the learning process and the parameters of the first intermediate layer in the first information processing device 2.

[0061] On the other hand, when the model provider learns, a learning process is carried out using the second input layer, intermediate layer group, and second output layer of the inference model, which are provided in the second information processing device 3, which is the information processing device on the model provider's side, and the first output layer and the first intermediate layer in the first information processing device 2 complete the learning process by reusing the parameters related to the second output layer determined by the learning process and the parameters related to the first intermediate layer in the second information processing device 3.

[0062] By configuring the information processing system 1 in this way, in addition to ensuring the confidentiality of the inference model, the confidentiality of medical data can be ensured because neither the correct answer data nor the learning data constituting the teacher data need to be transmitted to an external device. Also, the provider of the model can verify whether the learning process by the user of the inference model and the learning process by the provider of the inference model are being performed appropriately by inferring the inference target data for verifying the model using the second input layer, the intermediate layer group, and the second output layer.

[0063] The learning flow by the information processing system 1 in this embodiment will be described below with reference to FIG.

[0064] In step S1000, the teacher data acquisition unit 41 acquires teacher data from the storage unit 40, and the process proceeds to the next step.

[0065] In step S1001, the learning unit determines a path for executing an inference process in the inference model depending on whether the acquisition of teacher data or the input of the learning data constituting the teacher data is an input to the first input layer of the first partial model constituting the first information processing device 2 or an input to the second input layer of the second partial model constituting the second information processing device 3. Specifically, when the learning data is input to the first input layer, a learning process is performed using a first path for performing an inference process using the intermediate layer group and the first output layer, and when the learning data is input to the second input layer, a learning process is performed using a second path for performing an inference process using the intermediate layer group and the second output layer. That is, when the learning data is input to the first input layer, the learning unit advances the process to step S1002, and when the learning data is input to the second input layer, the learning unit advances the process to step S1004.

[0066] In step S1002, the learning unit forward propagates the learning data to the first input layer, the first hidden layer, the second hidden layer, and the first output layer. Also, the learning unit obtains loss information between the output from the first output layer and the correct answer data, and determines parameters for each layer by backpropagation based on the loss information.

[0067] In step S1003, the parameters related to the first output layer are copied and reused as the parameters related to the second output layer, and the process ends.

[0068] In step S1004, the learning unit forward propagates the learning data to the second input layer, the first hidden layer, the second hidden layer, and the second output layer. Also, the learning unit obtains loss information between the output from the second output layer and the correct answer data, and determines parameters for each layer by backpropagation based on the loss information.

[0069] In step S1004, the parameters related to the second output layer are copied and reused as the parameters related to the first output layer, and the process is terminated. Note that the parameters related to the first output layer and the second output layer may not be copies of the parameters related to one output layer, but may be determined by learning on the side of one information processing device using loss information calculated by the other information processing device.

[0070] By configuring the information processing system 1 in this way, in addition to ensuring the confidentiality of the inference model, the confidentiality of medical data can be ensured because neither the correct answer data nor the learning data constituting the teacher data need to be transmitted to an external device. Also, the provider of the model can verify whether the learning process by the user of the inference model and the learning process by the provider of the inference model are being performed appropriately by inferring the inference target data for verifying the model using the second input layer, the intermediate layer group, and the second output layer.

[0071] (Variation 2) In the learning process in the second and fourth embodiments, the case where the learning process of the inference model is performed using the backpropagation method has been described.

[0072] In this modified example, a case will be described in which the learning process for an inference model is performed using a learning method other than the backpropagation method.

[0073] For example, techniques such as Synthetic Gradient, which trains a model to estimate the gradient that will be obtained for each layer, Feedback Alignment, which uses a fixed random matrix when backpropagating errors, Target Prop, which propagates a target instead of an error, or any other technique are acceptable.

[0074] (Other Examples) The present invention can also be realized by executing the following process: That is, software (programs) that realize the functions of the above-described embodiments are supplied to a system or device via a network or various storage media, and the computer (or CPU, MPU, etc.) of the system or device reads and executes the programs.

Claims

1. a first information processing device and a second information processing device capable of communicating with the first information processing device via a network; An information processing system that performs a predetermined inference process using an inference model based on a neural network including a first input layer, a group of intermediate layers, a first output layer, and a second output layer, wherein the first output layer and the second output layer are provided in different information processing devices, the first information processing device has a first partial model including the first input layer, a first intermediate layer that is at least a part of the intermediate layers of the intermediate layer group, and the first output layer; the second information processing device has a second partial model configured to include a second intermediate layer in the intermediate layer group, the second intermediate layer including an intermediate layer different from an intermediate layer constituting the first intermediate layer, and the second output layer; an inference target data acquisition unit that acquires inference target data; an inference unit that performs a predetermined inference process on the inference target data by using the first partial model and the second partial model; An information processing system comprising:

2. The information processing system according to claim 1, characterized in that the inference unit performs the predetermined inference processing on the inference target data input to the first input layer using the intermediate layer group, the first output layer, and the second output layer.

3. the neural network further comprises a second input layer; The information processing system according to claim 1 , wherein the second partial model further includes the second input layer.

4. The inference unit When the inference target data is input to the first input layer, performing a predetermined inference process on the inference target data using at least the first input layer, the first hidden layer, the second hidden layer, and the first output layer; When the inference target data is input to the second input layer, 4. The information processing system according to claim 3, wherein a predetermined inference process is performed using at least the second input layer, the first intermediate layer, the second intermediate layer, and the second output layer.

5. a first information processing device and a second information processing device capable of communicating with the first information processing device via a network; An information processing system that performs a learning process to learn an inference model based on a neural network composed of a first input layer, a group of intermediate layers, a first output layer, and a second output layer, wherein the first output layer and the second output layer are provided in different information processing devices, the first information processing device has a first partial model including the first input layer, a first intermediate layer that is at least a part of the intermediate layers of the intermediate layer group, and the first output layer; the second information processing device has a second partial model configured to include a second intermediate layer in the intermediate layer group, the second intermediate layer including an intermediate layer different from an intermediate layer constituting the first intermediate layer, and the second output layer; a teacher data acquisition unit that acquires teacher data; a learning unit that uses the training data to learn the first partial model and the second partial model; An information processing system comprising:

6. The learning unit a learning unit that inputs learning data constituting the teacher data into the first input layer of the inference model and learns the intermediate layer group and the first output layer, The information processing system described in claim 5, characterized in that the intermediate layer group and the first output layer are trained by an error backpropagation method using loss information calculated using the correct answer data constituting the training data and the first output layer, and the parameters related to the first output layer obtained by the training are used as parameters related to the second output layer, thereby training the inference model.

7. the neural network further comprises a second input layer; The information processing system according to claim 5 , wherein the second partial model further includes the second input layer.

8. When learning data constituting the teacher data is input to the first input layer, the learning unit: The intermediate layer group and the first output layer are trained by a backpropagation method using loss information calculated using the correct answer data constituting the training data and the first output layer, and the parameters related to the first output layer obtained by the training are used as the parameters related to the second output layer. The inference model is learned by using When the training data constituting the teacher data is input to the second input layer, The intermediate layer group and the second output layer are trained by error backpropagation using loss information calculated using the correct answer data constituting the training data and the second output layer, and the parameters related to the second output layer obtained by the training are used as the parameters related to the first output layer.

8. The information processing system according to claim 7, wherein the inference model is learned by using the inference model as a parameter.

9. a first information processing device and a second information processing device capable of communicating with the first information processing device via a network; An information processing system that performs at least one of inference and learning using an inference model based on a neural network configured with a first input layer, a second input layer, a group of intermediate layers, a first output layer, and a second output layer, wherein the first input layer and the second input layer are provided in different information processing devices, The information processing system includes: An information processing system characterized by performing at least one of inference and learning on the inference model in a path corresponding to an input layer to which target data is input.

10. 10. The information processing system according to claim 9, wherein the paths include a first path using the first input layer, the group of hidden layers, and the first output layer, and a second path using the second input layer, the group of hidden layers, and the second output layer.

11. 11. The information processing system according to claim 9, wherein the first output layer and the second output layer are provided in different information processing devices.

12. the first input layer and the first output layer are provided in the same information processing device, 12. The information processing system according to claim 11, wherein the second input layer and the second output layer are provided in the same information processing device.

Images