Towards a General Framework for Continual Learning with Pre-training

TL;DR

This paper introduces a comprehensive framework for continual learning using pre-training, combining theoretical insights and practical techniques like PEFT to improve task adaptation and inference in dynamic environments.

Contribution

It presents a novel hierarchical framework for continual learning, integrating theoretical decomposition and parameter-efficient fine-tuning methods, with empirical validation.

Findings

Outperforms existing methods in downstream continual learning tasks

Demonstrates the effectiveness of PEFT in upstream continual learning

Provides a biological perspective linking neuroscience and continual learning

Abstract

In this work, we present a general framework for continual learning of sequentially arrived tasks with the use of pre-training, which has emerged as a promising direction for artificial intelligence systems to accommodate real-world dynamics. From a theoretical perspective, we decompose its objective into three hierarchical components, including within-task prediction, task-identity inference, and task-adaptive prediction. Then we propose an innovative approach to explicitly optimize these components with parameter-efficient fine-tuning (PEFT) techniques and representation statistics. We empirically demonstrate the superiority and generality of our approach in downstream continual learning, and further explore the applicability of PEFT techniques in upstream continual learning. We also discuss the biological basis of the proposed framework with recent advances in neuroscience.