Feng Zhu

Research

My research interests lie broadly in sequential decision-making under uncertainty, including online learning (e.g., stochastic bandits) [2, 3, 5, 6] and online matching (e.g., resource allocation) [1, 4, 7, 8], with applications to online experimentation, supply chain management and revenue management. My research goal is to develop data-driven decision-making paradigms that ensure the safety & resiliency of modern operational systems. In particular, I’m focusing on:

  • Managing (hidden) risk in various decision-making environments:
    • tail risk control for single bandit experiment [5, 6] [Job Market Paper];
    • false discovery control for multiple hypothesis testing [7];
    • risk detection and mitigation for supply chain resiliency [8].
  • Understanding the robustness of celebrated policies towards (structured) non-stationary environments with (extra) uncertainties:
    • if they work well with little or no amendment, what are the performance guarantees [1, 2, 3, 4];
    • if they may fail, what are the causes and how to fix them via new policy designs [5, 6, 7, 8].

Practically, I have been working with several large companies (Accenture, DENSO, Ford) to help design and implement supply chain risk detection models and mitigation strategies. Part of my work appeared in Harvard Business Review: Fixing the U.S. Semiconductor Supply Chain. In Summer 2023, I spent a great time working on inventory simulation and optimization as a supply chain analytics intern at Ford Motor Company.

Publications and Preprints (sorted in reverse chronological order; *represents α-β author order)

[9] Dynamic Service Fee Pricing under Strategic Behavior: Actions as Instruments and Phase Transition. Rui Ai*, David Simchi-Levi*, Feng Zhu*. Under preparation.

  • Preliminary version to appear in NeurIPS 2024.
  • INFORMS 2024.

[8] Risk Detection and Inventory Coordination under Uncertain Time-To-Recover. Pengfeng Shu*, David Simchi-Levi*, Chung-Piaw Teo*, Feng Zhu*. (One-page abstract. Full draft available upon request.)

  • INFORMS 2024.

[7] Bayesian Online Multiple Testing: A Resource Allocation Approach. ​Ruicheng Ao*, Hongyu Chen*, David Simchi-Levi*, Feng Zhu*.

  • Finalist, RMP 2024 Jeff McGill Best Student Paper Award.
  • INFORMS 2024, NUS Next-Gen Scholars Symposium, Purdue Operations Conference 2024, MSOM 2024.

[6] Regret Distribution in Stochastic Bandits: Optimal Trade-off between Expectation and Tail Risk. David Simchi-Levi*, Zeyu Zheng*, Feng Zhu*.

  • Preliminary version appeared in NeurIPS 2023 Spotlight (top 3%).
  • Finalist, POMS-HK 2024 Best Student Paper Competition.
  • INFORMS 2024, MSOM 2024, POMS 2024, POMS-HK 2024, INFORMS 2023, Purdue Operations Conference 2023.

[5] A Simple and Optimal Policy Design with Safety against Heavy-Tailed Risk for Stochastic Bandits. David Simchi-Levi*, Zeyu Zheng*, Feng Zhu*.

[4] On Greedy-like Policies in Online Matching with Reusable Network Resources and Decaying Rewards. David Simchi-Levi*, Zeyu Zheng*, Feng Zhu*.

  • Accepted by Management Science.
  • INFORMS 2022, Marketplace Innovation Workshop 2022.

[3] Offline Planning and Online Learning under Recovering Rewards. David Simchi-Levi*, Zeyu Zheng*, Feng Zhu*.

[2] Dynamic Pricing in a Non-stationary Growing Environment. Feng Zhu, Zeyu Zheng. 

[1] Assign-to-Seat: Dynamic Capacity Control for Selling High-Speed Train Tickets. Feng Zhu, Shaoxuan Liu, Rowan Wang and Zizhuo Wang.

  • Published in Manufacturing & Service Operations Management.