Some Considerations on Learning to Explore via Meta-Reinforcement Learning

TL;DR

This paper introduces two new meta-reinforcement learning algorithms, E-MAML and E-RL2, designed to improve exploration efficiency, demonstrated through experiments on novel environments and maze tasks.

Contribution

The paper proposes two novel meta-reinforcement learning algorithms specifically aimed at enhancing exploration capabilities.

Findings

E-MAML and E-RL2 outperform existing methods in exploration tasks

Demonstrated effectiveness on 'Krazy World' and maze environments

Improved performance in environments requiring strategic exploration

Abstract

We consider the problem of exploration in meta reinforcement learning. Two new meta reinforcement learning algorithms are suggested: E-MAML and E-$\text{RL}^2$. Results are presented on a novel environment we call `Krazy World' and a set of maze environments. We show E-MAML and E-$\text{RL}^2$ deliver better performance on tasks where exploration is important.