Look where you look! Saliency-guided Q-networks for generalization in visual Reinforcement Learning

TL;DR

This paper introduces saliency-guided Q-networks (SGQN), a novel approach that enhances visual reinforcement learning policies' ability to generalize across input disturbances by focusing on important pixels, leading to improved performance and interpretability.

Contribution

The paper proposes SGQN, a new method that guides value function learning with saliency information, significantly improving generalization and interpretability in visual RL tasks.

Findings

SGQN outperforms existing methods on the Deepmind Control Generalization benchmark.

SGQN improves training efficiency and reduces the generalization gap.

SGQN enhances policy interpretability by identifying important pixels.

Abstract

Deep reinforcement learning policies, despite their outstanding efficiency in simulated visual control tasks, have shown disappointing ability to generalize across disturbances in the input training images. Changes in image statistics or distracting background elements are pitfalls that prevent generalization and real-world applicability of such control policies. We elaborate on the intuition that a good visual policy should be able to identify which pixels are important for its decision, and preserve this identification of important sources of information across images. This implies that training of a policy with small generalization gap should focus on such important pixels and ignore the others. This leads to the introduction of saliency-guided Q-networks (SGQN), a generic method for visual reinforcement learning, that is compatible with any value function learning method. SGQN vastly improves the generalization capability of Soft Actor-Critic agents and outperforms existing stateof-the-art methods on the Deepmind Control Generalization benchmark, setting a new reference in terms of training efficiency, generalization gap, and policy interpretability.