Action abstractions for amortized sampling

TL;DR

This paper introduces a method for incorporating high-level action abstractions into reinforcement learning policies, improving exploration and sample efficiency in complex environments by chunking common action subsequences.

Contribution

It presents a novel iterative approach to extract and incorporate action abstractions into RL, enabling hierarchical planning and better exploration in long-horizon tasks.

Findings

Enhanced sample efficiency in discovering high-reward objects.

Abstracted actions are interpretable and reveal latent reward structure.

Improved performance on challenging exploration problems.

Abstract

As trajectories sampled by policies used by reinforcement learning (RL) and generative flow networks (GFlowNets) grow longer, credit assignment and exploration become more challenging, and the long planning horizon hinders mode discovery and generalization. The challenge is particularly pronounced in entropy-seeking RL methods, such as generative flow networks, where the agent must learn to sample from a structured distribution and discover multiple high-reward states, each of which take many steps to reach. To tackle this challenge, we propose an approach to incorporate the discovery of action abstractions, or high-level actions, into the policy optimization process. Our approach involves iteratively extracting action subsequences commonly used across many high-reward trajectories and `chunking' them into a single action that is added to the action space. In empirical evaluation on synthetic and real-world environments, our approach demonstrates improved sample efficiency performance in discovering diverse high-reward objects, especially on harder exploration problems. We also observe that the abstracted high-order actions are interpretable, capturing the latent structure of the reward landscape of the action space. This work provides a cognitively motivated approach to action abstraction in RL and is the first demonstration of hierarchical planning in amortized sequential sampling.