Timezone: »

Hindsight Learning for MDPs with Exogenous Inputs
Sean R. Sinclair · Felipe Vieira Frujeri · Ching-An Cheng · Luke Marshall · Hugo Barbalho · Jingling Li · Jennifer Neville · Ishai Menache · Adith Swaminathan

Tue Jul 25 02:00 PM -- 04:30 PM (PDT) @ Exhibit Hall 1 #602

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.

Author Information

Sean R. Sinclair (Cornell University)
Felipe Vieira Frujeri (Microsoft)
Ching-An Cheng (Microsoft Research)
Luke Marshall
Hugo Barbalho (Research, Microsoft)
Jingling Li (University of Maryland, College Park)
Jennifer Neville (Purdue University)
Ishai Menache (Technion - Israel Institute of Technology, Technion)
Adith Swaminathan (Microsoft Research)

More from the Same Authors