Meta few-shot class incremental learning

Patent metadata

Field	Value
Publication number	US-11983240-B2
Application number	US-202117527295-A
Country	US
Kind code	B2
Filing date	Nov 16, 2021
Priority date	Nov 16, 2021
Publication date	May 14, 2024
Grant date	May 14, 2024

Abstract

This disclosure provides for methods and system for meta few-shot class incremental learning. According to an aspect a method is provided. The method includes obtaining at least one weight attention map of a first network and updating weights of a second network using the at least one weight attention map, where the second network is a modulatory network. The method further includes generating at least one feature attention map of the second network based on the at least one weight attention map of the first network and a set of input images of at least one class. The method further includes generating at least one feature map of the first network based on the set of input images of the at least one class, and updating the at least one feature map of the first network based on the feature attention map of the second network.

First claim

What is claimed is:

1. A method comprising:
obtaining at least one weight attention map of a first network;
updating weights of a second network using the at least one weight attention map of the first network, wherein the second network is a modulatory network;
generating at least one feature attention map of the second network based on the at least one weight attention map of the first network and a set of input images of at least one class;
generating at least one feature map of the first network based on the set of input images of the at least one class; and
updating the at least one feature map of the first network based on the feature attention map of the second network.

2. The method of claim 1 , wherein the first network is a prediction network.

3. The method of claim 1 , wherein the updating weights of the second network comprises multiplying the weight of the second network with the at least one weight attention map.

4. The method of claim 1 , wherein the at least one weight attention map represents a current state of the first network, the current state indicating a learnt knowledge.

5. The method of claim 1 further comprising:
providing to the first network a set of training data indicative of n novel classes;
adding n nodes to a classifier of the first network, wherein the classifier comprises one or more existing nodes indicative of one or more learnt classes of the at least one class, each of the existing node indicative of a learnt class;
merging, via concatenation, the n nodes with the one or more existing nodes;
updating the first network based on the set of training data; and
evaluating the updated first network based on test data indicative of the novel classes and learnt classes.

6. The method of claim 5 , wherein the set of training data and the set of test data are from a set of sequential data.

7. The method of claim 6 , wherein the set of sequential data and the set if input images are from a set of base data indicative of base classes, bases classes comprising the n novel classes and the at least one class.

8. The method of claim 1 wherein each of the first network and the second network is a deep neural network.

9. The method of claim 5 , wherein the n novel classes is from a different domain of classes than the at least one class.

10. A method comprising:
providing to a network a set of training data indicative of n novel classes;
adding n nodes to a classifier of the network, wherein the classifier comprises one or more existing nodes indicative of one or more learnt classes, each of the existing node indicative of a learnt class;
merging, via concatenation, the n nodes with the one or more existing nodes;
updating the network based on the set of training data; and
evaluating the updated network based on test data indicative of the novel classes and learnt classes.

11. The method of claim 10 , wherein the network is a prediction network.

12. The method of claim 10 , wherein the set of training data represents an n-way-k-shot classification problem.

13. The method of claim 10 , wherein the evaluating comprises: generating a loss function for the updated prediction network based on the test data.

14. The method of claim 10 wherein the updating the prediction network comprises:
updating weight parameters of a backbone of the prediction network based on the set of training data.

15. The method of claim 13 further comprising:
updating weight parameters of a backbone of the prediction network based on the generated loss function.

16. A device comprising at least one processor and at least one non-transitory machine-readable medium storing executable instructions which when executed by the at least one processor configure the device for:
obtaining at least one weight attention map of a first network;
updating weights of a second network using the at least one weight attention map of the first network, wherein the second network is a modulatory network;
generating at least one feature attention map of the second network based on the at least one weight attention map of the first network and a set of input images of at least one class;
generating at least one feature map of the first network based on the set of input images of the at least one class; and
updating the at least one feature map of the first network based on the feature attention map of the second network.

17. The device of claim 16 , wherein the at least one weight attention map represents a current state of the first network, the current state indicating a learnt knowledge.

18. The device of claim 16 , wherein the at least one processor further configures the device for:
providing to the first network a set of training data indicative of n novel classes;
adding n nodes to a classifier of the first network, wherein the classifier comprises one or more existing nodes indicative of one or more learnt classes of the at least one class, each of the existing node indicative of a learnt class;
merging, via concatenation, the n nodes with the one or more existing nodes;
updating the first network based on the set of training data; and
evaluating the updated first network based on test data indicative of the novel classes and learnt classes.

19. The device of claim 18 , wherein the evaluating comprises: generating a loss function for the updated prediction network based on the test data.

20. The device of claim 19 , wherein the at least one processor further configures the device for updating weight parameters of a backbone of the prediction network based on the generated loss function.

Assignees

• Huawei Tech Co Ltd

Inventors

• Chi Zhixiang
• Gu Li
• Liu Huan
• Yu Yuanhao
• Wang Yang
• Tang Jin

Classifications

• G06F18/213 Primary

  Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods · CPC title

• G06F18/2148

  characterised by the process organisation or structure, e.g. boosting cascade · CPC title

• G06F18/217

  Validation; Performance evaluation; Active pattern learning techniques · CPC title

• G06F18/243

  relating to the number of classes · CPC title

• G06N3/045

  Combinations of networks · CPC title