Progress & Compress: A scalable framework for continual learning

TL;DR

The paper presents Progress & Compress, a scalable continual learning framework that preserves past knowledge while efficiently learning new tasks without increasing model size or requiring previous data, demonstrated on classification and reinforcement learning tasks.

Contribution

It introduces a novel continual learning method that maintains performance on prior tasks without architecture growth or data storage, using a cycle of active learning and consolidation.

Findings

Effective on sequential alphabet classification

Performs well on Atari games

Succeeds in 3D maze navigation

Abstract

We introduce a conceptually simple and scalable framework for continual learning domains where tasks are learned sequentially. Our method is constant in the number of parameters and is designed to preserve performance on previously encountered tasks while accelerating learning progress on subsequent problems. This is achieved by training a network with two components: A knowledge base, capable of solving previously encountered problems, which is connected to an active column that is employed to efficiently learn the current task. After learning a new task, the active column is distilled into the knowledge base, taking care to protect any previously acquired skills. This cycle of active learning (progression) followed by consolidation (compression) requires no architecture growth, no access to or storing of previous data or tasks, and no task-specific parameters. We demonstrate the progress & compress approach on sequential classification of handwritten alphabets as well as two reinforcement learning domains: Atari games and 3D maze navigation.