# US 2026/0065059 A1 · Classification apparatus, classification method, and classification program

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

- Number: US 2026/0065059 A1
- Document ID: US20260065059A1
- Status: Published application
- Kind: A1
- Application number: 19/314,790
- Filed: 2025-08-29
- Published: 2026-03-05
- Earliest claimed: 2024-09-02
- Assignees: JVCKENWOOD Corporation
- Applicants: JVCKENWOOD Corporation
- Inventors: Yincheng YANG
- Web page: https://patentgrep.com/patent/US20260065059A1

## CPC

- G06N 3/082: modifying the architecture, e.g. adding, deleting or silencing nodes or connections
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/09: Supervised learning
- G06N 3/096: Transfer learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A classification apparatus includes: a feature quantity output unit that outputs a feature quantity of input data; and a classification unit that retains, as a classification weight, a feature quantity obtained by averaging, per each class, the feature quantity output by the feature quantity output unit in response to a base class dataset and a novel class dataset with a smaller number of data items than the base class dataset and that outputs a result of classification of the input data by using the feature quantity of the input data and the classification weight.

## Claims

### Claim 1

A classification apparatus comprising:
a feature quantity output unit that outputs a feature quantity of input data; and
a classification unit that retains, as a classification weight, a feature quantity obtained by averaging, per each class, the feature quantity output by the feature quantity output unit in response to a base class dataset and a novel class dataset with a smaller number of data items than the base class dataset and that outputs a result of classification of the input data by using the feature quantity of the input data and the classification weight,
wherein the feature quantity output unit is generated by subjecting a neural network to training that uses the base class dataset, the training including removing or adding a path across nodes between adjacent layers in the neural network and then training the neural network by distillation by using the novel class dataset with reference to a further feature quantity output unit as a supervisor model.

### Claim 2 (depends on 1)

The classification apparatus according claim 1,
wherein the classification unit retains, as the classification weight, a feature quantity obtained by adding and averaging, per each class, a feature quantity output by the further feature quantity output unit in response to the novel class dataset as an input to the feature quantity output by the feature quantity output unit in response to base class data and novel class data as inputs.

### Claim 3 (depends on 1)

The classification apparatus according claim 1, wherein the further feature quantity output unit includes:
a neural network trained by using the base class dataset and outputting a feature quantity of the input data;
a scaling unit that adjusts a value of the feature quantity output by the neural network by multiplying a multiplication value by the feature quantity; and
a bias unit that adds an addition value to the value adjusted by the scaling unit,
wherein the multiplication value and the addition value are updated by inner learning that uses a support set for the base class,
wherein the inner learning is performed by a learning apparatus that includes the further feature quantity output unit, a further classification unit, and a learning unit,
wherein, in the inner learning,
the neural network outputs the feature quantity in response to the support set for the base class as an input,
the scaling unit outputs a multiplication result obtained by multiplying the multiplication value by the feature quantity output by the neural network,
the bias unit outputs an addition result obtained by adding the addition value to the multiplication result output by the scaling unit,
the further classification unit retains a condensed classification weight which is a weight for classification into each class and outputs a classification result from the addition result and the condensed classification weight in response to the addition result output by the bias unit as an input, and
the learning unit calculates a loss in response to the classification result output by the further classification unit as an input and updates the multiplication value and the addition value based on the loss.

### Claim 4 (depends on 3)

The classification apparatus according claim 3,
wherein the condensed classification weight is updated by outer learning that uses a query set for the base class after the multiplication value and the addition value are updated, wherein the outer learning is performed by the learning apparatus,
wherein, in the outer learning,
the neural network outputs the feature quantity in response to the query set for the base class as an input,
the scaling unit outputs a multiplication result obtained by multiplying the multiplication value by the feature quantity output by the neural network,
the bias unit outputs an addition result obtained by adding the addition value to the multiplication result output by the neural network,
the further classification unit retains a condensed classification weight which is a weight for classification into each class and outputs a classification result from the addition result and the condensed classification weight in response to the addition result output by the bias unit as an input, and
the learning unit calculates a loss in response to the classification result output by the further classification unit as an input and updates the condensed classification weight based on the loss.

### Claim 5

A classification method comprising:
outputting a feature quantity of input data; and
retaining, as a classification weight, a feature quantity obtained by averaging, per each class, the feature quantity output by the outputting in response to a base class dataset and a novel class dataset with a smaller number of data items than the base class dataset and outputting a result of classification of the input data by using the feature quantity of the input data and the classification weight,
wherein a feature quantity output unit executing the outputting is generated by being subject to training using the base class dataset, the training including removing or adding a path across nodes between adjacent layers in a neural network and then being trained by distillation by using the novel class dataset with reference to a further feature quantity output unit as a supervisor model.

### Claim 6

A classification program comprising computer-implemented modules including:
a feature quantity output module that outputs a feature quantity of input data; and
a classification module that retains, as a classification weight, a feature quantity obtained by averaging, per each class, the feature quantity output by the feature quantity output module in response to a base class dataset and a novel class dataset with a smaller number of data items than the base class dataset and that outputs a result of classification of the input data by using the feature quantity of the input data and the classification weight,
wherein a feature quantity output unit executing the feature quantity output module is generated by being subject to training using the base class dataset, the training including removing or adding a path across nodes between adjacent layers in a neural network and then being trained by distillation by using the novel class dataset with reference to a further feature quantity output unit as a supervisor model.
