DeCoR: Defy Knowledge Forgetting by Predicting Earlier Audio Codes
This addresses the problem of adapting audio models to new data over time without forgetting old tasks, which is incremental as it builds on existing continual learning methods.
The paper tackles catastrophic forgetting in lifelong audio feature extraction by introducing DeCoR, a method that indirectly distills knowledge from earlier models to the latest by predicting quantization indices from a delayed codebook, improving acoustic scene classification accuracy with minimal storage and computation overhead.
Lifelong audio feature extraction involves learning new sound classes incrementally, which is essential for adapting to new data distributions over time. However, optimizing the model only on new data can lead to catastrophic forgetting of previously learned tasks, which undermines the model's ability to perform well over the long term. This paper introduces a new approach to continual audio representation learning called DeCoR. Unlike other methods that store previous data, features, or models, DeCoR indirectly distills knowledge from an earlier model to the latest by predicting quantization indices from a delayed codebook. We demonstrate that DeCoR improves acoustic scene classification accuracy and integrates well with continual self-supervised representation learning. Our approach introduces minimal storage and computation overhead, making it a lightweight and efficient solution for continual learning.