Generalization in Reinforcement Learning with Selective Noise Injection and Information Bottleneck

TL;DR

This paper introduces Selective Noise Injection and combines it with the Information Bottleneck to improve the generalization ability of reinforcement learning policies, especially in low-data regimes, outperforming existing methods on benchmarks.

Contribution

It proposes a novel adaptation of regularization techniques for RL, specifically Selective Noise Injection, and demonstrates the effectiveness of combining it with the Information Bottleneck.

Findings

SNI maintains regularization benefits while improving gradient quality.

Combining IB with SNI outperforms state-of-the-art on Coinrun.

The approach enhances generalization in RL, especially early in training.

Abstract

The ability for policies to generalize to new environments is key to the broad application of RL agents. A promising approach to prevent an agent's policy from overfitting to a limited set of training environments is to apply regularization techniques originally developed for supervised learning. However, there are stark differences between supervised learning and RL. We discuss those differences and propose modifications to existing regularization techniques in order to better adapt them to RL. In particular, we focus on regularization techniques relying on the injection of noise into the learned function, a family that includes some of the most widely used approaches such as Dropout and Batch Normalization. To adapt them to RL, we propose Selective Noise Injection (SNI), which maintains the regularizing effect the injected noise has, while mitigating the adverse effects it has on the gradient quality. Furthermore, we demonstrate that the Information Bottleneck (IB) is a particularly well suited regularization technique for RL as it is effective in the low-data regime encountered early on in training RL agents. Combining the IB with SNI, we significantly outperform current state of the art results, including on the recently proposed generalization benchmark Coinrun.