Control method and information processing apparatus

Patent metadata

Field	Value
Publication number	US-12315221-B2
Application number	US-202217830171-A
Country	US
Kind code	B2
Filing date	Jun 1, 2022
Priority date	Jan 17, 2020
Publication date	May 27, 2025
Grant date	May 27, 2025

Abstract

A storage unit holds a classification model that calculates a confidence score from image data, and a transformation model that is a model for transforming a feature value having fewer dimensions than the image data into the image data and is created such that a set of feature values corresponding to a set of image data follows a probability distribution. A processing unit extracts a feature value according to the probability distribution. The processing unit transforms the feature value into image data using the transformation model and calculates a confidence score corresponding to the image data using the classification model. The processing unit updates, based on the probability distribution and the feature value, a feature value to be input to the transformation model from the feature value to a feature value in such a manner that a confidence score to be calculated is higher than the confidence score.

First claim

What is claimed is:

1. A control method comprising:
obtaining, by a processor, a classification model and a transformation model, the classification model being configured to calculate, from input image data, a confidence score indicating a likelihood that the input image data belongs to a specified class, the transformation model being a model for transforming an input feature value having fewer dimensions than the input image data into the input image data and being created such that a set of feature values corresponding to a set of image data is included in a feature space having a specific probability distribution;
extracting, by the processor, a first feature value from the feature space;
transforming, by the processor, the first feature value into first image data using the transformation model, and calculating a first confidence score corresponding to the first image data using the classification model; and
calculating, by the processor, a weight coefficient using a first probability density corresponding to the first feature value, and searching for a second feature value from the feature space based on a product of the first confidence score and the weight coefficient in such a manner that a second confidence score to be calculated by the classification model from second image data corresponding to the second feature value is higher than the first confidence score, the first probability density being given by the specific probability distribution.

2. The control method according to claim 1 , further comprising creating, by the processor, a variational autoencoder including an encoder and a decoder, and using the decoder as the transformation model, the encoder being configured to transform the input image data into the input feature value, the decoder being configured to transform the input feature value into the input image data.

3. The control method according to claim 1 , further comprising:
setting, by the processor, an upper limit value that is lower than a maximum probability density value of the specific probability distribution,
wherein the searching further includes using the first probability density as the weight coefficient in response to the first probability density being less than or equal to the upper limit value, and using the upper limit value as the weight coefficient in response to the first probability density being greater than the upper limit value.

4. The control method according to claim 1 , wherein:
the extracting includes extracting a plurality of first feature values including the first feature value;
the searching includes updating the plurality of first feature values to a plurality of second feature values including the second feature value; and
the control method further includes
calculating, by the processor, an evaluation value of each of the plurality of second feature values, based on a second confidence score calculated from the each of the plurality of second feature values, and
selecting, by the processor, at least one second feature value from the plurality of second feature values, based on the evaluation value of the each of the plurality of second feature values, extracting an additional second feature value existing within a predetermined range from the at least one second feature value, and adding the additional second feature value as the feature value to be input to the transformation model.

5. An information processing apparatus comprising:
a memory configured to hold a classification model and a transformation model, the classification model being configured to calculate, from input image data, a confidence score indicating a likelihood that the input image data belongs to a specified class, the transformation model being a model for transforming an input feature value having fewer dimensions than the input image data into the input image data and being created such that a set of feature values corresponding to a set of image data is included in a feature space having a specific probability distribution; and
a processor configured to
extract a first feature value from the feature space,
transform the first feature value into first image data using the transformation model, and calculate a first confidence score corresponding to the first image data using the classification model, and
calculate a weight coefficient using a first probability density corresponding to the first feature value, and search for a second feature value from the feature space based on a product of the first confidence score and the weight coefficient in such a manner that a second confidence score to be calculated by the classification model from second image data corresponding to the second feature value is higher than the first confidence score, the first probability density being given by the specific probability distribution.

6. A control method comprising:
extracting, by a processor, a first value from a feature space, the feature space including a plurality of values based on features of a plurality of image datasets and having a specific distribution, the plurality of values each having fewer variables than each of the plurality of image datasets;
obtaining, by the processor, a first confidence score of a specified group among confidence scores of groups included in a classification inference result obtained by a classification inference model receiving the first value as an input value; and
calculating, by the processor, a weight coefficient using a first probability density corresponding to the first value, and searching for a second value from the feature space based on a product of the first confidence score and the weight coefficient such that a second confidence score of the specified group to be included in a classification inference result obtained by the classification inference model receiving the second value as an input value is higher than the first confidence score, the first probability density being given by the specific distribution.

Assignees

• Fujitsu Ltd

Inventors

• Higuchi Yuji
• Shimizu Toshiya
• Morikawa Ikuya

Classifications

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/04

  Architecture, e.g. interconnection topology · CPC title

• G06V10/776

  Validation; Performance evaluation · CPC title