Learning to Search Better Than Your Teacher

TL;DR

The paper introduces LOLS, a novel learning to search algorithm that not only performs well relative to a reference policy but also guarantees low regret against deviations from the learned policy, enabling improvements over suboptimal references.

Contribution

LOLS provides a new learning to search method with local-optimality guarantees, allowing it to outperform poor reference policies and facilitating structured contextual bandits.

Findings

LOLS guarantees low regret compared to deviations from the learned policy.

LOLS can improve upon suboptimal reference policies.

Develops structured contextual bandits for partial information settings.

Abstract

Methods for learning to search for structured prediction typically imitate a reference policy, with existing theoretical guarantees demonstrating low regret compared to that reference. This is unsatisfactory in many applications where the reference policy is suboptimal and the goal of learning is to improve upon it. Can learning to search work even when the reference is poor? We provide a new learning to search algorithm, LOLS, which does well relative to the reference policy, but additionally guarantees low regret compared to deviations from the learned policy: a local-optimality guarantee. Consequently, LOLS can improve upon the reference policy, unlike previous algorithms. This enables us to develop structured contextual bandits, a partial information structured prediction setting with many potential applications.