The Importance Of Sampling InMeta-Reinforcement Learning

Authors:
Bradly Stadie Vector Institute
Ge Yang Berkeley
Rein Houthooft Happy Elements
Peter Chen covariant.ai
Yan Duan UC Berkeley
Yuhuai Wu University of Toronto
Pieter Abbeel UC Berkeley / Gradescope / Covariant
Ilya Sutskever OpenAI

Introduction:

The authors interpret meta-reinforcement learning as the problem of learning how to quickly find a good sampling distribution in a new environment.

Abstract:

We interpret meta-reinforcement learning as the problem of learning how to quickly find a good sampling distribution in a new environment. This interpretation leads to the development of two new meta-reinforcement learning algorithms: E-MAML and E-$\text{RL}^2$. Results are presented on a new environment we call `Krazy World': a difficult high-dimensional gridworld which is designed to highlight the importance of correctly differentiating through sampling distributions in meta-reinforcement learning. Further results are presented on a set of maze environments. We show E-MAML and E-$\text{RL}^2$ deliver better performance than baseline algorithms on both tasks.

You may want to know: