Beyond Black-Box Advice: Learning-Augmented Algorithms for MDPs with Q-Value Predictions

TL;DR

This paper introduces learning-augmented algorithms for MDPs that leverage Q-value advice with known generation processes, achieving a novel balance between consistency and robustness and improving performance over black-box advice methods.

Contribution

It presents the first tradeoff analysis for Q-value advice in MDPs with known advice generation, enabling dynamic selection between advice and robust baselines.

Findings

Achieves near-optimal performance guarantees.

Demonstrates improved tradeoff over black-box advice.

Applicable to both continuous and discrete MDPs.

Abstract

We study the tradeoff between consistency and robustness in the context of a single-trajectory time-varying Markov Decision Process (MDP) with untrusted machine-learned advice. Our work departs from the typical approach of treating advice as coming from black-box sources by instead considering a setting where additional information about how the advice is generated is available. We prove a first-of-its-kind consistency and robustness tradeoff given Q-value advice under a general MDP model that includes both continuous and discrete state/action spaces. Our results highlight that utilizing Q-value advice enables dynamic pursuit of the better of machine-learned advice and a robust baseline, thus result in near-optimal performance guarantees, which provably improves what can be obtained solely with black-box advice.