Few-Shot Learning Device Selective Base Class Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In incremental few-shot learning, the convergence of loss is difficult and time-consuming due to the need to project all pre-learned base classes onto a joint classification space for meta-learning, which increases learning time and decreases classification accuracy by considering irrelevant classes in the loss calculation.
Innovation Solution
A machine learning device and method that selectively projects a predetermined number of base classes onto a projection space during meta-learning, reducing the number of classification target classes and focusing on highly relevant classes for loss calculation, thereby facilitating loss convergence and improving classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all pre-learned base classes are projected onto a joint classification space for meta-learning, then comprehensive classification coverage is achieved, but learning time increases and classification accuracy decreases due to considering irrelevant classes
Solution Approach 1:
The patent segments the classification space by separating base classes into two groups: those that are projected onto the joint classification space and those that are not. This segmentation allows the system to focus computational resources on relevant classes while excluding irrelevant ones, thereby improving classification accuracy and reducing learning time simultaneously.
Solution Approach 2:
The patent extracts and excludes irrelevant base classes from the joint classification space projection. By taking out classes that are not relevant to the current task, the system avoids the computational overhead and accuracy degradation caused by considering all base classes, thus resolving the contradiction between comprehensive coverage and efficiency.
2Adaptability or versatility
If all pre-learned base classes are projected onto a joint classification space for meta-learning, then comprehensive classification coverage is achieved, but device complexity increases
Solution Approach 1:
The patent segments the set of base classes into relevant and irrelevant subsets, projecting only the relevant ones onto the joint classification space. This segmentation reduces the complexity of the learning system by limiting the number of classes that need to be processed, while still maintaining adaptability for the current task through selective inclusion.
Solution Approach 2:
The patent applies local quality by treating different base classes differently based on their relevance to the current task. Relevant classes receive full projection and processing, while irrelevant classes are excluded, thereby optimizing the balance between classification coverage and system complexity through differentiated treatment of class subsets.
Data Source
AI summary
A machine learning device is provided that performs continual learning of a fewer number of novel classes than the number of base classes. A base class feature extraction unit extracts feature vectors of the base classes. A novel class feature extraction unit extracts feature vectors of the novel classes. A mixture feature calculation unit mixes the feature vectors of the base classes and the feature vectors of the novel classes and calculates a mixture feature vector of the base classes and the novel classes. A learning unit classifies a query sample of a query set based on the distance between the position of a mixture feature vector of the query sample of the query set and the position of a classification weight vector of each class in a projection space and learns classification weight vectors of the novel classes so as to minimize classification loss.