Think Too Fast Nor Too Slow: The Computational Trade-off Between Planning And Reinforcement Learning

TL;DR

This paper investigates the computational trade-off between planning and reinforcement learning, revealing that optimal performance occurs at a balance point rather than at extremes of planning duration.

Contribution

It introduces the importance of balancing planning and learning time, and conceptualizes a spectrum of algorithms from exhaustive search to model-free RL.

Findings

Optimal performance is achieved with a balanced approach to planning and learning.

Long or short planning durations are suboptimal for decision-making.

A new spectrum of planning-learning algorithms is proposed.

Abstract

Planning and reinforcement learning are two key approaches to sequential decision making. Multi-step approximate real-time dynamic programming, a recently successful algorithm class of which AlphaZero [Silver et al., 2018] is an example, combines both by nesting planning within a learning loop. However, the combination of planning and learning introduces a new question: how should we balance time spend on planning, learning and acting? The importance of this trade-off has not been explicitly studied before. We show that it is actually of key importance, with computational results indicating that we should neither plan too long nor too short. Conceptually, we identify a new spectrum of planning-learning algorithms which ranges from exhaustive search (long planning) to model-free RL (no planning), with optimal performance achieved midway.