SLCA: Slow Learner with Classifier Alignment for Continual Learning on a Pre-trained Model

TL;DR

This paper introduces SLCA, a simple yet effective method for continual learning on pre-trained models that addresses overfitting by slow learning and classifier alignment, significantly improving performance across various datasets.

Contribution

The paper proposes SLCA, a novel approach combining slow learning in representation layers with classifier alignment, to enhance continual learning on pre-trained models.

Findings

Achieves up to 49.76% improvement on Split CIFAR-100

Outperforms state-of-the-art methods by large margins

Provides a strong baseline for future research in CLPM

Abstract

The goal of continual learning is to improve the performance of recognition models in learning sequentially arrived data. Although most existing works are established on the premise of learning from scratch, growing efforts have been devoted to incorporating the benefits of pre-training. However, how to adaptively exploit the pre-trained knowledge for each incremental task while maintaining its generalizability remains an open question. In this work, we present an extensive analysis for continual learning on a pre-trained model (CLPM), and attribute the key challenge to a progressive overfitting problem. Observing that selectively reducing the learning rate can almost resolve this issue in the representation layer, we propose a simple but extremely effective approach named Slow Learner with Classifier Alignment (SLCA), which further improves the classification layer by modeling the class-wise distributions and aligning the classification layers in a post-hoc fashion. Across a variety of scenarios, our proposal provides substantial improvements for CLPM (e.g., up to 49.76%, 50.05%, 44.69% and 40.16% on Split CIFAR-100, Split ImageNet-R, Split CUB-200 and Split Cars-196, respectively), and thus outperforms state-of-the-art approaches by a large margin. Based on such a strong baseline, critical factors and promising directions are analyzed in-depth to facilitate subsequent research. Code has been made available at: https://github.com/GengDavid/SLCA.