Maintaining Plasticity in Continual Learning via Regenerative Regularization

TL;DR

This paper introduces L2 Init, a straightforward regularization method that helps neural networks retain plasticity in continual learning by nudging parameters toward their initial values, thereby improving adaptation to new data.

Contribution

L2 Init is a simple, effective regularization technique that maintains neural network plasticity in continual learning by regularizing toward initial parameters.

Findings

L2 Init consistently mitigates plasticity loss across various non-stationary tasks.

L2 Init outperforms previous methods in maintaining adaptability.

The method is easy to implement with only one hyper-parameter.

Abstract

In continual learning, plasticity refers to the ability of an agent to quickly adapt to new information. Neural networks are known to lose plasticity when processing non-stationary data streams. In this paper, we propose L2 Init, a simple approach for maintaining plasticity by incorporating in the loss function L2 regularization toward initial parameters. This is very similar to standard L2 regularization (L2), the only difference being that L2 regularizes toward the origin. L2 Init is simple to implement and requires selecting only a single hyper-parameter. The motivation for this method is the same as that of methods that reset neurons or parameter values. Intuitively, when recent losses are insensitive to particular parameters, these parameters should drift toward their initial values. This prepares parameters to adapt quickly to new tasks. On problems representative of different types of nonstationarity in continual supervised learning, we demonstrate that L2 Init most consistently mitigates plasticity loss compared to previously proposed approaches.