I'm an undergraduate researcher at Shandong University. I build empirical, reproducible research at the intersection of machine learning, optimization, and decision-making under uncertainty.
My current work focuses on offline reinforcement learning, robust supply-chain design, and understanding when learned representations help—or fail—in downstream decisions.
Diagnosing regime representations in offline reinforcement learning
An empirical study of whether learned dataset-regime representations improve offline RL under heterogeneous behavior data. The project separates classification quality from downstream utility and benchmarks learned features against semantic and non-semantic controls across Hopper, Walker2d, and HalfCheetah.
Python · PyTorch · Offline RL · IQL · D4RL · Representation Learning
Comparing hurricane-risk representations in two-stage stochastic network design
A CVaR-based supply-chain design study comparing static, rule-based, interpretable probabilistic, and Random-Forest risk models. It builds an event-node dataset from 464 North Atlantic hurricanes and validates reduced-scenario designs against the full scenario set.
Python · scikit-learn · Gurobi · Stochastic Optimization · CVaR
- Offline and reinforcement learning
- Representation learning and diagnostic evaluation
- Stochastic and robust optimization
- Data-driven decision-making under uncertainty
Python · PyTorch · scikit-learn · Gurobi · Jupyter · Git