On the Effectiveness of LayerNorm Tuning for Continual Learning in Vision Transformers

TL;DR

This paper proposes tuning LayerNorm parameters for continual learning in Vision Transformers, reducing computational costs while maintaining competitive performance through a novel two-stage training and inference selection method.

Contribution

It introduces a simple yet effective method of learning task-specific LayerNorm parameters for continual learning, improving efficiency without sacrificing accuracy.

Findings

Achieves state-of-the-art or comparable results on ImageNet-R and CIFAR-100.

Reduces computational costs compared to existing rehearsal-free methods.

Demonstrates robustness to incorrect parameter selection during inference.

Abstract

State-of-the-art rehearsal-free continual learning methods exploit the peculiarities of Vision Transformers to learn task-specific prompts, drastically reducing catastrophic forgetting. However, there is a tradeoff between the number of learned parameters and the performance, making such models computationally expensive. In this work, we aim to reduce this cost while maintaining competitive performance. We achieve this by revisiting and extending a simple transfer learning idea: learning task-specific normalization layers. Specifically, we tune the scale and bias parameters of LayerNorm for each continual learning task, selecting them at inference time based on the similarity between task-specific keys and the output of the pre-trained model. To make the classifier robust to incorrect selection of parameters during inference, we introduce a two-stage training procedure, where we first optimize the task-specific parameters and then train the classifier with the same selection procedure of the inference time. Experiments on ImageNet-R and CIFAR-100 show that our method achieves results that are either superior or on par with {the state of the art} while being computationally cheaper.