Semi-Discriminative Representation Loss for Online Continual Learning

The use of episodic memory in continual learning has demonstrated effectiveness for alleviating catastrophic forgetting. In recent studies, gradient-based approaches have been developed to make more efficient use of compact episodic memory. Such approaches refine the gradients resulting from new sam...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2022-04
Main Authors: Chen, Yu, Diethe, Tom, Flach, Peter
Format: Article
Language:English
Subjects:
Initiatives
Learning
Optimization
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The use of episodic memory in continual learning has demonstrated effectiveness for alleviating catastrophic forgetting. In recent studies, gradient-based approaches have been developed to make more efficient use of compact episodic memory. Such approaches refine the gradients resulting from new samples by those from memorized samples, aiming to reduce the diversity of gradients from different tasks. In this paper, we clarify the relation between diversity of gradients and discriminativeness of representations, showing shared as well as conflicting interests between Deep Metric Learning and continual learning, thus demonstrating pros and cons of learning discriminative representations in continual learning. Based on these findings, we propose a simple method -- Semi-Discriminative Representation Loss (SDRL) -- for continual learning. In comparison with state-of-the-art methods, SDRL shows better performance with low computational cost on multiple benchmark tasks in the setting of online continual learning.
ISSN:2331-8422