Dimension-Wise Importance Sampling Weight Clipping for Sample-Efficient Reinforcement Learning

TL;DR

This paper introduces a dimension-wise importance sampling weight clipping method for PPO that reduces bias and enhances sample efficiency in high-dimensional action spaces, outperforming existing algorithms.

Contribution

It proposes a novel dimension-wise IS weight clipping technique for PPO, improving learning efficiency and sample reuse in high-dimensional action tasks.

Findings

Outperforms PPO in OpenAI Gym tasks

Reduces bias in high-dimensional action spaces

Enables effective sample reuse

Abstract

In importance sampling (IS)-based reinforcement learning algorithms such as Proximal Policy Optimization (PPO), IS weights are typically clipped to avoid large variance in learning. However, policy update from clipped statistics induces large bias in tasks with high action dimensions, and bias from clipping makes it difficult to reuse old samples with large IS weights. In this paper, we consider PPO, a representative on-policy algorithm, and propose its improvement by dimension-wise IS weight clipping which separately clips the IS weight of each action dimension to avoid large bias and adaptively controls the IS weight to bound policy update from the current policy. This new technique enables efficient learning for high action-dimensional tasks and reusing of old samples like in off-policy learning to increase the sample efficiency. Numerical results show that the proposed new algorithm outperforms PPO and other RL algorithms in various Open AI Gym tasks.