Learning Action Representations for Reinforcement Learning
This addresses a bottleneck in model-free RL for domains with large action spaces, offering a novel method to improve generalization.
The paper tackles the problem of generalization over large action sets in reinforcement learning by learning action representations, which allow agents to infer outcomes of similar actions, and demonstrates efficacy on large-scale real-world problems.
Most model-free reinforcement learning methods leverage state representations (embeddings) for generalization, but either ignore structure in the space of actions or assume the structure is provided a priori. We show how a policy can be decomposed into a component that acts in a low-dimensional space of action representations and a component that transforms these representations into actual actions. These representations improve generalization over large, finite action sets by allowing the agent to infer the outcomes of actions similar to actions already taken. We provide an algorithm to both learn and use action representations and provide conditions for its convergence. The efficacy of the proposed method is demonstrated on large-scale real-world problems.