InfBA: Interference-Free Bottleneck Adaptation for Continual Learning.

Abstract

Continual learning requires the model to learn multiple tasks sequentially. In continual learning, the model should possess the ability to maintain its performance on old tasks (stability) and the ability to adapt to new tasks continuously (plasticity). Recently, parameter-efficient fine-tuning (PEFT), which involves freezing a pre-trained model and injecting a small number of learnable parameters to adapt to downstream tasks, has gained increasing popularity in continual learning. Although existing continual learning methods based on PEFT have demonstrated superior performance compared to those not based on PEFT, most of them do not consider how to eliminate the interference of the new task on the old tasks, which inhibits the model from making a good trade-off between stability and plasticity. In this work, we propose a new PEFT method, called interference-free bottleneck adaptation (InfBA), for continual learning. InfBA adopts a bottleneck architecture, which decreases the dimensionality of the embedding first and then increases it. Since bottleneck architecture has been utilized by many existing PEFT methods such as Adapter, LoRA and Prefix-tuning, InfBA provides a framework to integrate with these methods. InfBA constrains the update within a subspace, and designs this subspace to eliminate the interference of the new task on the old tasks, making a good trade-off between stability and plasticity. Experimental results on multiple datasets show that our methods consistently outperform existing state-of-the-art continual learning methods based on PEFT.