# US 2021/0241021 A1 · Information processing method and information processing system

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0241021 A1
- Document ID: US20210241021A1
- Status: Published application
- Kind: A1
- Application number: 17/234,127
- Filed: 2021-04-19
- Published: 2021-08-05
- Earliest claimed: 2019-03-14
- Assignees: Panasonic Intellectual Property Corporation of America
- Applicants: Panasonic Intellectual Property Corporation of America
- Inventors: Denis GUDOVSKIY; Alec HODGKINSON; Takuya YAMAGUCHI; Sotaro TSUKIZAWA
- Web page: https://patentgrep.com/patent/US20210241021A1

## CPC

- G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/09: Supervised learning
- G06N 3/091: Active learning
- G06N 5/04: Inference or reasoning models
- G06N 20/00: Machine learning
- G06N 3/045: Combinations of networks
- G06N 3/08: Learning methods
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

An information processing method implemented by a computer includes: obtaining a piece of first data, and a piece of second data not included in a training dataset for training an inferencer; calculating, using a piece of first relevant data obtained by inputting the first data to the inferencer trained by machine learning using the training dataset, a first contribution representing contributions of portions constituting the first data to a piece of first output data output by inputting the first data to the inferencer; calculating, using a piece of second relevant data obtained by inputting the second data to the inferencer, a second contribution representing contributions of portions constituting the second data to a piece of second output data output by inputting the second data to the inferencer; and determining whether to add the second data to the training dataset, according to the similarity between the first and second contributions.

## Claims

### Claim 1

An information processing method implemented by a computer, the information processing method comprising:
obtaining a piece of first data and a piece of second data, the piece of second data being not included in a training dataset used for training an inferencer;
calculating, by using a piece of first relevant data, a first contribution representing respective contributions of portions constituting the piece of first data to a piece of first output data output by inputting the piece of first data to the inferencer, the piece of first relevant data being obtained by inputting the piece of first data to the inferencer trained by machine learning using the training dataset;
calculating, by using a piece of second relevant data, a second contribution representing respective contributions of portions constituting the piece of second data to a piece of second output data output by inputting the piece of second data to the inferencer, the piece of second relevant data being obtained by inputting the piece of second data to the inferencer; and
determining whether to add the piece of second data to the training dataset for the inferencer, according to a degree of similarity between the first contribution and the second contribution.

### Claim 2 (depends on 1)

The information processing method according to claim 1,
wherein the piece of first data is a piece of data for which the inferencer made a false inference.

### Claim 3 (depends on 1)

The information processing method according to claim 1,
wherein if a determination to add the piece of second data is made, the inferencer is trained using the training dataset to which the piece of second data has been added, the training dataset being used for training the inferencer.

### Claim 4 (depends on 1)

The information processing method according to claim 1,
wherein in the determining of whether to add the piece of second data to the training dataset for the inferencer,
whether to add to the training dataset for the inferencer is determined for each of pieces of second data to enable a piece of second data having a higher degree of similarity between the first contribution and the second contribution to be more preferentially selected from the pieces of second data and added to the training dataset for the inferencer, the pieces of second data each being the piece of second data.

### Claim 5 (depends on 1)

The information processing method according to claim 1,
wherein in the determining of whether to add the piece of second data to the training dataset for the inferencer,
a plurality of the first contributions that include the first contribution of each of pieces of first data are calculated, the pieces of first data each being the piece of first data, and
whether to add the piece of second data to the training dataset for the inferencer is determined according to a representative value calculated using a plurality of degrees of similarity including degrees of similarity between the plurality of the first contributions and the second contribution.

### Claim 6 (depends on 5)

The information processing method according to claim 5,
wherein the plurality of degrees of similarity include degrees of similarity between the plurality of the first contributions and a plurality of the second contributions that include the second contribution of each of pieces of second data, the pieces of second data each being the piece of second data,
regarding obtainment of the representative value,
for each of the pieces of first data, a predetermined number of degrees of similarity are selected from the plurality of degrees of similarity, and
for each of the pieces of second data, the representative value is calculated by using the predetermined number of degrees of similarity, and
in the determining of whether to add the piece of second data to the training dataset for the inferencer,
whether to add the piece of second data to the training dataset for the inferencer is determined according to the representative values.

### Claim 7 (depends on 1)

The information processing method according to claim 1,
wherein the inferencer is an identifier or a detector.

### Claim 8 (depends on 1)

The information processing method according to claim 1,
wherein each of the piece of first data and the piece of second data is a piece of sensing data.

### Claim 9 (depends on 1)

The information processing method according to claim 1,
wherein the piece of first relevant data is the piece of first output data, and
the piece of second relevant data is the piece of second output data.

### Claim 10 (depends on 1)

The information processing method according to claim 1,
wherein a presentation apparatus presents information indicating the piece of second data determined to be added to the training dataset for the inferencer.

### Claim 11

An information processing system, comprising:
an acquisition unit that obtains a piece of first data and a piece of second data, the piece of second data being not included in a training dataset used for training an inferencer;
a computation unit (a) that calculates, by using a piece of first relevant data, a first contribution representing respective contributions of portions constituting the piece of first data to a piece of first output data output by inputting the piece of first data to the inferencer, the piece of first relevant data being obtained by inputting the piece of first data to the inferencer trained by machine training using the training dataset and (b) that calculates, by using a piece of second relevant data, a second contribution representing respective contributions of portions constituting the piece of second data to a piece of second output data output by inputting the piece of second data to the inferencer, the piece of second relevant data being obtained by inputting the piece of second data to the inferencer; and
a determination unit that determines whether to add the piece of second data to the training dataset for the inferencer, according to a degree of similarity between the first contribution and the second contribution.
