Research

Central to my research are tools from numerical linear algebra, economics and mathematical optimization. My research mission is to develop Efficient Algorithmic Tools to enable scalable and socially responsible operations in modern complex systems such as digital market, healthcare and criminal justice, and so on.

Selective publications

† equal contribution

Computation of Competitive Equilibrium
Equilibrium computation is a central step to justify long-run outcomes of any decision-making process. This series of work brings interior-point methods to reduce the computational effort for classical general equilibrium settings in economics.
The second-order tâtonnement: decentralized interior-point methods for market equilibrium
Chuwen Zhang, Chang He, Bo Jiang, Yinyu Ye
Breaking the traditional belief on centralization and information burden that discourages high-order methods.
Major Revision at Operations Research, {arXiv}
Tractable approximation of Arrow-Debreu equilibrium by interior-point price adjustments
An approximation to bypass PPAD-hardness.
Chuwen Zhang, Chang He, Yinyu Ye
in preparation, available upon request
Homogeneous Framework for Second-Order Methods
My PhD thesis builds on work that uses symmetric eigenvalue problems in place of Newton systems for highly degenerate problems.
A homogeneous second-order descent method for nonconvex optimization
Chuwen Zhang, Chang He, Yuntian Jiang, Dongdong Ge, Bo Jiang, Yinyu Ye
Math. Oper. Res. · {arXiv}
Homogeneous second-order descent framework: a fast alternative to Newton-type methods
Chang He†, Yuntian Jiang†, Chuwen Zhang†, Dongdong Ge, Bo Jiang, Yinyu Ye
Math. Prog. · {arXiv}
Applications: linear systems, and a policy gradient method for reinforcement learning (published on UAI).
Optimal Treatment Control for Diversion Programs in Criminal Justice
Optimal rehabilitation policies for diversion programs, with theoretical guarantees on long-run performance.
The dynamic and endogenous behavior of re-offense risk: an agent-based simulation study of treatment allocation in incarceration diversion programs
Chuwen Zhang, Pengyi Shi, Amy Ward
Part I: discrete-event simulation framework for dynamic re-offense risk under treatment.
{arXiv}
Better resource allocations in the criminal justice system: optimizing for the long-term good
Chuwen Zhang, Pengyi Shi, Amy Ward
Part II: optimal long-run treatment allocation with provable guarantees.
in preparation
$\bullet$ Media coverage: the post "Can AI Reduce the Prison Population?" on Chicago Booth Review.
Misc.
During my time at Cardinal Operations, I had some experience in optimization solver development, e.g., linear conic solvers like cuPDLP-C, ABIP (on IJOC). These efforts have become the basis of modern GPU-based linear conic solvers in COPT, Gurobi, and NVIDIA.
I was also responsible of a few interesting industrial applications; see the research reports on aircraft engine maintenance problem (on EJOR), Jinghu high-speed railway scheduling (on TPAMI).