Dynamically Modular and Sparse General Continual Learning

TL;DR

This paper introduces Dynamos, a dynamic modular and sparse neural network approach inspired by brain sparse coding, to improve continual learning by reducing task interference and catastrophic forgetting.

Contribution

The paper proposes Dynamos, a novel method that dynamically activates relevant neuron subsets, enhancing modularity and sparsity in continual learning scenarios.

Findings

Dynamos effectively reduces catastrophic forgetting.

The method learns modular, specialized representations.

It maintains reusability through overlapping neuron activations.

Abstract

Real-world applications often require learning continuously from a stream of data under ever-changing conditions. When trying to learn from such non-stationary data, deep neural networks (DNNs) undergo catastrophic forgetting of previously learned information. Among the common approaches to avoid catastrophic forgetting, rehearsal-based methods have proven effective. However, they are still prone to forgetting due to task-interference as all parameters respond to all tasks. To counter this, we take inspiration from sparse coding in the brain and introduce dynamic modularity and sparsity (Dynamos) for rehearsal-based general continual learning. In this setup, the DNN learns to respond to stimuli by activating relevant subsets of neurons. We demonstrate the effectiveness of Dynamos on multiple datasets under challenging continual learning evaluation protocols. Finally, we show that our method learns representations that are modular and specialized, while maintaining reusability by activating subsets of neurons with overlaps corresponding to the similarity of stimuli.