Knowledge Transfer in Deep Reinforcement Learning via an RL-Specific GAN-Based Correspondence Function

TL;DR

This paper presents a novel RL-specific GAN-based method for effective one-to-one knowledge transfer in deep reinforcement learning, improving generalization and reducing training time in complex tasks.

Contribution

It introduces a modified Cycle GAN with new loss components tailored for RL, enabling superior knowledge transfer compared to standard GANs.

Findings

Achieved 100% knowledge transfer in identical tasks.

Reduced training time by 30% in rotated tasks.

Outperformed standard GANs in transfer performance.

Abstract

Deep reinforcement learning has demonstrated superhuman performance in complex decision-making tasks, but it struggles with generalization and knowledge reuse - key aspects of true intelligence. This article introduces a novel approach that modifies Cycle Generative Adversarial Networks specifically for reinforcement learning, enabling effective one-to-one knowledge transfer between two tasks. Our method enhances the loss function with two new components: model loss, which captures dynamic relationships between source and target tasks, and Q-loss, which identifies states significantly influencing the target decision policy. Tested on the 2-D Atari game Pong, our method achieved 100% knowledge transfer in identical tasks and either 100% knowledge transfer or a 30% reduction in training time for a rotated task, depending on the network architecture. In contrast, using standard Generative Adversarial Networks or Cycle Generative Adversarial Networks led to worse performance than training from scratch in the majority of cases. The results demonstrate that the proposed method ensured enhanced knowledge generalization in deep reinforcement learning.