When Waiting is not an Option : Learning Options with a Deliberation Cost

TL;DR

This paper introduces a framework for learning options in reinforcement learning that balances performance and interpretability by incorporating a deliberation cost, inspired by bounded rationality, with demonstrated improvements in the Arcade Learning Environment.

Contribution

It formulates a novel approach to learning options based on deliberation costs within the bounded rationality framework and provides practical gradient-based algorithms for implementation.

Findings

Enhanced performance in Arcade Learning Environment

Improved interpretability of learned options

Effective incorporation of deliberation cost in option learning

Abstract

Recent work has shown that temporally extended actions (options) can be learned fully end-to-end as opposed to being specified in advance. While the problem of "how" to learn options is increasingly well understood, the question of "what" good options should be has remained elusive. We formulate our answer to what "good" options should be in the bounded rationality framework (Simon, 1957) through the notion of deliberation cost. We then derive practical gradient-based learning algorithms to implement this objective. Our results in the Arcade Learning Environment (ALE) show increased performance and interpretability.