Overcoming catastrophic forgetting in neural networks

TL;DR

This paper presents a scalable method to prevent catastrophic forgetting in neural networks by selectively slowing down learning on important weights, enabling sequential learning of multiple tasks like MNIST classification and Atari games.

Contribution

The authors introduce a novel approach that maintains performance on old tasks while learning new ones by adjusting learning rates based on task importance.

Findings

Effective in preventing forgetting on MNIST and Atari tasks

Scalable approach suitable for complex sequential learning

Maintains high accuracy on previously learned tasks

Abstract

The ability to learn tasks in a sequential fashion is crucial to the development of artificial intelligence. Neural networks are not, in general, capable of this and it has been widely thought that catastrophic forgetting is an inevitable feature of connectionist models. We show that it is possible to overcome this limitation and train networks that can maintain expertise on tasks which they have not experienced for a long time. Our approach remembers old tasks by selectively slowing down learning on the weights important for those tasks. We demonstrate our approach is scalable and effective by solving a set of classification tasks based on the MNIST hand written digit dataset and by learning several Atari 2600 games sequentially.