Towards Redundancy-Free Sub-networks in Continual Learning

TL;DR

This paper introduces IBM, a novel method for continual learning that creates redundancy-free sub-networks by leveraging information bottleneck principles, significantly reducing parameters and training time while mitigating catastrophic forgetting.

Contribution

IBM is the first approach to eliminate redundancy within sub-networks in continual learning, improving efficiency and effectiveness over existing parameter isolation methods.

Findings

IBM reduces sub-network parameters by 70%.

IBM decreases training time by 80%.

IBM outperforms state-of-the-art methods in continual learning.

Abstract

Catastrophic Forgetting (CF) is a prominent issue in continual learning. Parameter isolation addresses this challenge by masking a sub-network for each task to mitigate interference with old tasks. However, these sub-networks are constructed relying on weight magnitude, which does not necessarily correspond to the importance of weights, resulting in maintaining unimportant weights and constructing redundant sub-networks. To overcome this limitation, inspired by information bottleneck, which removes redundancy between adjacent network layers, we propose \textbf{\underline{I}nformation \underline{B}ottleneck \underline{M}asked sub-network (IBM)} to eliminate redundancy within sub-networks. Specifically, IBM accumulates valuable information into essential weights to construct redundancy-free sub-networks, not only effectively mitigating CF by freezing the sub-networks but also facilitating new tasks training through the transfer of valuable knowledge. Additionally, IBM decomposes hidden representations to automate the construction process and make it flexible. Extensive experiments demonstrate that IBM consistently outperforms state-of-the-art methods. Notably, IBM surpasses the state-of-the-art parameter isolation method with a 70\% reduction in the number of parameters within sub-networks and an 80\% decrease in training time.