When Waiting is not an Option : Learning Options with a Deliberation Cost
This work addresses a fundamental challenge in reinforcement learning for AI agents by providing a principled way to learn options, though it is incremental as it builds on existing end-to-end learning methods.
The paper tackled the problem of defining 'good' temporally extended actions (options) in reinforcement learning by introducing deliberation cost within a bounded rationality framework, and demonstrated improved performance and interpretability in the Arcade Learning Environment.
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.