In behavior economics, a decision maker’s (DM) preference over prospects can be expressed by choice functions (e.g., utility function and risk measure). In this work, we study preference robust optimization (PRO) problems where robust optimal decision should be made with ambiguous choice function of DM and its partial information can be elicited. We show novel ways of defining and formulating multi-attribute quasi-concave choice functions. Mathematical tractability schemes and new decision analysis foundations are illustrated. Efficient numerical methods are developed where the robust choice function and the PRO problem can be constructed and solved by a sequence of linear programs/convex optimization problems. Finally, we test the behavior and scalability of our method numerically on a portfolio optimization problem and a capital allocation problem.
Bio: Wenjie Huang is currently an international postdoctoral fellow at School of Data Science, The Chinese University of Hong Kong, Shenzhen and Group for Research in Decision Analysis, HEC Montreal, supervised by Zizhuo Wang and Erick Delage. He obtained Ph.D. degree from National University of Singapore in Jun 2019 and obtained BSc. degree from Shanghai Jiao Tong University in 2014. His research interests include: stochastic/robust optimization, reinforcement learning, data-driven decision-making, with applications in risk management, operations management and smart city operations.
Huang_Wenjie_PRO_talk.pdf (2 MB)