Learning Representations for New Sound Classes With Continual Self-Supervised Learning

TL;DR

This paper introduces a continual self-supervised learning framework for sound recognition that effectively incorporates new classes with minimal labeled data, demonstrating robustness to forgetting and comparable performance to distillation methods.

Contribution

It proposes a novel continual learning approach using representation learning that is robust to forgetting and works with unlabeled data for sound classification.

Findings

Similarity-based representation learning is robust to forgetting.

The method performs comparably to distillation-based approaches.

Effective with minimal labeled data in a continual setting.

Abstract

In this paper, we work on a sound recognition system that continually incorporates new sound classes. Our main goal is to develop a framework where the model can be updated without relying on labeled data. For this purpose, we propose adopting representation learning, where an encoder is trained using unlabeled data. This learning framework enables the study and implementation of a practically relevant use case where only a small amount of the labels is available in a continual learning context. We also make the empirical observation that a similarity-based representation learning method within this framework is robust to forgetting even if no explicit mechanism against forgetting is employed. We show that this approach obtains similar performance compared to several distillation-based continual learning methods when employed on self-supervised representation learning methods.