PhD Dissertation Proposal by PHAM Hoang Giang | Optimization under Choice and Behavioral Response Models: Algorithms and Performance Guarantees

ABOUT THE TALK

Retailers deciding which products to offer and at what prices, and defenders allocating limited protection against boundedly rational attackers, must anticipate how people respond to their decisions. Logit-family choice models describe these responses well, but building them into optimization produces nonconvex and often NP-hard problems, which practice usually tackles with heuristics or local search that give no measure of the remaining error. This dissertation proposal shows that the difficulty has an identifiable structure, and that exposing it yields algorithms with provable performance guarantees. It answers three research questions, one for each kind of decision. Discrete assortment decisions admit exact methods and approximation guarantees. For continuous decisions such as pricing and security allocation, and for estimating the choice models themselves, the hard part can be confined to a few customer segments, attacker types, nests, or model parameters, so solutions can be certified as globally optimal within a stated tolerance. Decisions that mix binary and continuous variables admit approximations with an explicit bound on the loss. Eight manuscripts support the proposal, and their experiments show that the methods certify instances that standard solvers leave open.

ABOUT THE SPEAKER

PHAM Hoang Giang is a PhD candidate in Computer Science at the School of Computing and Information Systems, Singapore Management University, advised by Assistant Professor MAI Anh Tien. Giang's research is in decision making and optimization, with a focus on optimization under discrete choice and behavioral response models and its applications to assortment and pricing, facility location, security games, and the estimation of choice models. This work has appeared in the European Journal of Operational Research and the Proceedings of IJCAI 2026.