Online Resource Allocation with Personalized Learning

Abstract

Joint online learning and resource allocation is a fundamental problem inherent in many applications. In a general setting, heterogeneous customers arrive sequentially, each of which can be allocated to a resource in an online fashion. Customers stochastically consume the resources, allocations yield stochastic rewards, and the system receives feedback outcomes with delay. In “Online Resource Allocation with Personalized Learning,” Zhalechian, Keyvanshokooh, Shi, and Van Oyen introduce a generic framework to solve this problem. It judiciously synergizes online learning with a broad class of online resource allocation mechanisms, where the sequence of customer contexts is adversarial, and the customer reward and resource consumption are stochastic and unknown. They propose online algorithms that strike a three-way balance between exploration, exploitation, and hedging against adversarial arrival sequence. A performance guarantee is provided for each online algorithm, and the efficacy of their algorithms is demonstrated using clinical data from a health system.