Skip to main content

Thesis Information
Title: Analyzing Tatonnement Dynamics in Economics Markets
Adviser: Richard Cole
Institution: Courant Institute of Mathematical Sciences, New York University
Graduation Date: August 2014

Contact Information
Email Candidate
Candidate Website

Candidate Bio:

I have held postdoctoral positions in University of Vienna and Max Planck Institute (MPI) for Informatics, and currently in Singapore University of Technology and Design. Trained as a Theoretical Computer Science (TCS) researcher and having knowledge of Dynamical Systems (DS), I enjoy the interdisciplinary research at the intersection of Computer Science and Economics, where I have applied concepts and techniques in TCS and DS to analyze (in)stability of machine learning algorithm, asynchronous optimization and market dynamics. I also like thinking about problems in Combinatorics, particularly those arising from Graph Sparsification and Partitioning.

Research Summary:

"My research focuses on Computational Economics and Algorithmic Game Theory. Economies (e.g., games, markets, auctions) have agents making decisions (i.e., computing or learning) and interacting rapidly, with minimal centralized coordination. My long term goal is to develop a thorough understanding of complex economies from a modern computer-science perspective.

My interdisciplinary research in Computer Science and Economics has proved that the two subjects can gain insight from each other. Here are some highlights:

a) we use a volume expansion argument to show that Multiplicative Weights Update (MWU) and Follow-the-Regularized-Leader (FTRL) algorithms in zero-sum games are Lyapunov chaotic in a payoff space;

b) using duality and transformations of convex programs, we drew direct connections between market dynamics and gradient/mirror descent;

c) using amortized analysis and adversary construction, we derived tight bounds on permissible parallelism of Stochastic Asynchronous Coordinate Descent; we also showed a market dynamic is robust against asynchrony."

Teaching Aims:

"I enjoy teaching and mentoring students. It is always fulfilling to see students progress in my classes. Also, teaching improves myself; I view teaching as an indispensable process of developing myself into a better researcher and scholar.

My first teaching experience dates back to at least 17 years ago. In 2017, I was the lecturer in Max Planck Institute of a graduate level Algorithmic Game Theory course. I have developed my teaching philosophy with five principles:

a) emphasis of meaning and understanding;
b) try the very best to give motivation on the topic to be taught;
c) encourage critical thinking;
d) elicit potential of advanced students;
e) teaching and research are mutually beneficial to each other."

Paper 1:

Tatonnement Beyond Gross Substitutes? Gradient Descent to the Rescue (with Richard Cole and Nikhil Devanur), STOC'2013 and Games and Economics Behaviors 2019

Link to PDF

Paper 2:

Vortices Instead of Equilibria in MinMax Optimization: Chaos and Butterfly Effects of Online Learning in Zero-Sum Games (with Georgios Piliouras), COLT'2019

Link to PDF

Paper 3:

Steiner Point Removal -- Distant Terminals Don't (Really) Bother (single-author), SODA'2018

Link to PDF

Keywords: Computational Economics, Algorithmic Game Theory, Learning in Games, Asynchronous Optimization, Dynamical Systems

All Job Candidates