Model-Based Active Exploration

TL;DR

This paper presents MAX, a model-based active exploration algorithm that uses ensemble disagreement to efficiently guide exploration, significantly outperforming baselines in various environments.

Contribution

Introduces MAX, a novel active exploration method leveraging ensemble disagreement for planning, scalable to high-dimensional continuous environments.

Findings

MAX is at least ten times more efficient than baselines in semi-random environments.

MAX effectively scales to high-dimensional continuous tasks.

Ensemble disagreement serves as a useful measure of novelty for exploration.

Abstract

Efficient exploration is an unsolved problem in Reinforcement Learning which is usually addressed by reactively rewarding the agent for fortuitously encountering novel situations. This paper introduces an efficient active exploration algorithm, Model-Based Active eXploration (MAX), which uses an ensemble of forward models to plan to observe novel events. This is carried out by optimizing agent behaviour with respect to a measure of novelty derived from the Bayesian perspective of exploration, which is estimated using the disagreement between the futures predicted by the ensemble members. We show empirically that in semi-random discrete environments where directed exploration is critical to make progress, MAX is at least an order of magnitude more efficient than strong baselines. MAX scales to high-dimensional continuous environments where it builds task-agnostic models that can be used for any downstream task.