Incremental learning for audio classification with Hebbian Deep Neural Networks

TL;DR

This paper introduces a biologically inspired Hebbian learning approach for incremental sound classification, using kernel plasticity to improve stability and accuracy over multiple learning steps.

Contribution

It proposes a novel kernel plasticity method that selectively modulates network kernels during incremental learning, inspired by human lifelong learning.

Findings

Achieved 76.3% accuracy on ESC-50 over five incremental steps.

Outperformed baseline without kernel plasticity (68.7%).

Demonstrated greater stability across tasks.

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.