Center Loss Regularization for Continual Learning

TL;DR

This paper introduces a center loss regularization technique for continual learning that reduces catastrophic forgetting by maintaining class centers across tasks, improving performance with minimal overhead.

Contribution

The paper proposes a novel center loss regularization method that preserves class representations in sequential learning, outperforming existing replay strategies.

Findings

Outperforms other replay-based strategies in continual learning.

Effective on MNIST variants and domain adaptation datasets.

Requires minimal computational and memory resources.

Abstract

The ability to learn different tasks sequentially is essential to the development of artificial intelligence. In general, neural networks lack this capability, the major obstacle being catastrophic forgetting. It occurs when the incrementally available information from non-stationary data distributions is continually acquired, disrupting what the model has already learned. Our approach remembers old tasks by projecting the representations of new tasks close to that of old tasks while keeping the decision boundaries unchanged. We employ the center loss as a regularization penalty that enforces new tasks' features to have the same class centers as old tasks and makes the features highly discriminative. This, in turn, leads to the least forgetting of already learned information. This method is easy to implement, requires minimal computational and memory overhead, and allows the neural network to maintain high performance across many sequentially encountered tasks. We also demonstrate that using the center loss in conjunction with the memory replay outperforms other replay-based strategies. Along with standard MNIST variants for continual learning, we apply our method to continual domain adaptation scenarios with the Digits and PACS datasets. We demonstrate that our approach is scalable, effective, and gives competitive performance compared to state-of-the-art continual learning methods.