Generalized Adaptive Transfer Network: Enhancing Transfer Learning in Reinforcement Learning Across Domains

TL;DR

The paper introduces GATN, a novel deep reinforcement learning architecture that improves transfer learning across domains by enhancing generalization, robustness, and efficiency, demonstrated on multiple benchmarks.

Contribution

GATN is a new RL framework that incorporates domain-agnostic representations, robustness-aware policies, and efficient transfer scheduling for better cross-domain transfer.

Findings

GATN outperforms baselines in cross-domain generalization.

GATN shows increased resilience to environmental changes.

GATN reduces computational overhead in transfer learning.

Abstract

Transfer learning in Reinforcement Learning (RL) enables agents to leverage knowledge from source tasks to accelerate learning in target tasks. While prior work, such as the Attend, Adapt, and Transfer (A2T) framework, addresses negative transfer and selective transfer, other critical challenges remain underexplored. This paper introduces the Generalized Adaptive Transfer Network (GATN), a deep RL architecture designed to tackle task generalization across domains, robustness to environmental changes, and computational efficiency in transfer. GATN employs a domain-agnostic representation module, a robustness-aware policy adapter, and an efficient transfer scheduler to achieve these goals. We evaluate GATN on diverse benchmarks, including Atari 2600, MuJoCo, and a custom chatbot dialogue environment, demonstrating superior performance in cross-domain generalization, resilience to dynamic environments, and reduced computational overhead compared to baselines. Our findings suggest GATN is a versatile framework for real-world RL applications, such as adaptive chatbots and robotic control.