Hindsight Learning for MDPs with Exogenous Inputs

TL;DR

This paper introduces Hindsight Learning, a data-efficient approach for decision-making in Exo-MDPs with exogenous variables, demonstrating superior performance over existing methods in resource management tasks.

Contribution

The paper proposes a novel Hindsight Learning algorithm for Exo-MDPs that leverages exogenous variable samples to improve policy learning efficiency.

Findings

HL outperforms domain-specific heuristics.

HL surpasses state-of-the-art reinforcement learning methods.

Effective in real-world cloud resource management scenarios.

Abstract

Many resource management problems require sequential decision-making under uncertainty, where the only uncertainty affecting the decision outcomes are exogenous variables outside the control of the decision-maker. We model these problems as Exo-MDPs (Markov Decision Processes with Exogenous Inputs) and design a class of data-efficient algorithms for them termed Hindsight Learning (HL). Our HL algorithms achieve data efficiency by leveraging a key insight: having samples of the exogenous variables, past decisions can be revisited in hindsight to infer counterfactual consequences that can accelerate policy improvements. We compare HL against classic baselines in the multi-secretary and airline revenue management problems. We also scale our algorithms to a business-critical cloud resource management problem -- allocating Virtual Machines (VMs) to physical machines, and simulate their performance with real datasets from a large public cloud provider. We find that HL algorithms outperform domain-specific heuristics, as well as state-of-the-art reinforcement learning methods.