Incremental learning for audio classification with Hebbian Deep Neural Networks

2026-04-20

AI Summary

This work addresses catastrophic forgetting in continual learning for audio classification by proposing a Hebbian learning–based kernel plasticity method. The approach dynamically modulates convolutional kernels during incremental learning, selectively allocating subsets of kernels to acquire new classes while preserving others to retain previously learned knowledge. This mechanism effectively balances the model’s stability and plasticity. Evaluated on the ESC-50 dataset under a five-phase incremental learning protocol, the method achieves an overall accuracy of 76.3%, substantially outperforming a baseline without kernel plasticity, which attains only 68.7%. These results demonstrate the efficacy of the proposed strategy in simultaneously maintaining historical knowledge and adapting to new tasks.

Technology Category

• Machine Learning: Life-Long and Continual Learning
• Natural Language Processing: Learning & Optimization for NLP
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

The ability of humans for lifelong learning is an inspiration for deep learning methods and in particular for continual learning. In this work, we apply Hebbian learning, a biologically inspired learning process, to sound classification. We propose a kernel plasticity approach that selectively modulates network kernels during incremental learning, acting on selected kernels to learn new information and on others to retain previous knowledge. Using the ESC-50 dataset, the proposed method achieves 76.3% overall accuracy over five incremental steps, outperforming a baseline without kernel plasticity (68.7%) and demonstrating significantly greater stability across tasks.

Problem

• incremental learning
• audio classification
• continual learning
• catastrophic forgetting
• lifelong learning

Innovation

• Hebbian learning
• kernel plasticity
• incremental learning
• audio classification
• continual learning