I am a Ph.D. candidate in the Department of Industrial Engineering and Management Sciences at Northwestern University advised by Prof. Chang-Han Rhee.

I’m in the 2023-24 Academic Job Market. Please find my CV here. For a brief summary of my research, please find the recording of my lightning talk at the SNAPP seminar.

Upcoming Presentations

I will give a talk about “How to Characterize Global Dynamics with Rare-Event Analysis: A Heavy-Tail Framework in Deep Learning” over Zoom at the Lightning Talk Session at SNAPP Seminar:

• December 11 (Mon), 11:30 AM - 12:30 PM (Eastern Time), Zoom Link, SNAPP Seminar Website

I will give an advanced tutorial about “Importance Sampling Strategy for Heavy-Tailed Systems with Catastrophe Principle” at the 2023 Winter Simulation Conference:

• December 12 (Tue), 1:30 PM - 3:00 PM, Conference Room 8

I will present my poster “Large Deviations and Metastability Analysis for Heavy-Tailed Dynamical Systems” at the Heavy Tails in ML workshop in NeurIPS 2023:

• December 15 (Fri), 16:10 PM - 17:30 PM, Rooms R02-R05

Research

I’m broadly interested in applied probability, machine learning theory, and stochastic simulation. In particular, my work focuses on the heavy-tailed phenomenon (i.e., extreme variability) in stochastic systems. Contrary to common belief, heavy tail is often the norm rather than an exception in significant applications in areas such as finance, large-scale networks, and the training of AI. My work develops mathemaical tools such as large deviations to rigorously characterize how systems can deviate from expected behavior and how rare events manifest under the presence of heavy tails. What’s particularly surprising and valuable is that, based on knowledge of atypical behaviors (rare events), I can provide a comprehensive characterization of the typical behaviors (i.e., metastability) of the global dynamics (see [1]). Building upon such universal and powerful theories, I develop provably strong computational tools that can quantify, control, or even leverage the heavy-tailed phenomenon in modern algorithms and complex systems. For instance:

• My work [2] rigorously characterize a fascinating phenomenon where injecting and then truncating heavy-tailed noises during the training phase of deep neural networks can lead to improved generalization performance during the test phase.
• Our heavy-tailed large deviation theory can be applied to obtain a unified importance sampling scheme that is not only strongly efficient but also universarlly applicable to rare-event simulation tasks in a wide range of contexts; In [3] we illustrate its application in the contexts of option pricing, stochastic approximation, and queuing systems; In [4] we successfully extend the algorithm to address the particularly challenging cases where it is computationally infeasible to simulate or store the underlying process.

Honors and Awards

• Second Place, George Nicholson Student Paper Competition (2023), INFORMS
• Terminal Year Fellowship (2023), Northwestern University
• Nemhauser Prize for Best Student Paper (2022), IEMS, Northwestern University
• Benjamin A. Sachs Graduate Fellowship (2022), IEMS, Northwestern University
• Arthur P. Hurter Award for Academic Excellence among First Year Graduate Students (2019), IEMS, Northwestern University
• Lee Wai Wang Scholarship (2015), Peking University
• National Scholarship for Undergraduates (2013), Peking University