Haipeng Chen
William & Mary, VA·Computer Science
Many public health decisions come down to choosing the best subset from a large group. Examples include deciding which symptomatic people in a community get screened first, or which youth at a homeless shelter to invite to an educational program.
- Fields
Computer Science
Public Health & Epidemiology
Mathematics & Statistics
- Topics
reinforcement learning
combinatorial optimization
graph neural networks
- Grants
- 1 known grant, through Aug 2027
What they work on
- Develop physics-inspired, task-aware graph neural networks that build problem structure into reinforcement learning, so it can handle badly shaped or implicit objectives in combinatorial optimization.
- Design a robust reinforcement learning algorithm, based on adversarial training and game theory, that balances average-case and worst-case performance when problem parameters are uncertain.
- Create a hierarchical reinforcement learning algorithm that jointly learns high-level budget allocation and low-level selection of individuals for repeated (multi-shot) public health decisions.
More about the research
These are combinatorial optimization problems (COPs), and they are very hard to solve at real-world scale. Traditional solvers rely on hand-designed rules of thumb, which need expert knowledge and may give poor answers.
Why it matters
Better algorithms for choosing whom to screen, treat or enroll could help public health programs, shelters and conservation groups allocate limited resources and schedule tasks. They could make these decisions more effectively and on time. The methods aim to make reinforcement learning practical for real-world societal decisions where data are large and parameters are uncertain.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Pan Xu
Duke University, NC·Computer Science·Early-career
Develop distributionally robust reinforcement learning methods that use minimax optimization over a set of possible environment dynamics, so that policies trained in a simulator perform reliably in an unknown target domain, with finite-sample guarantees.
Craig Newgard
Oregon Health & Science University, OR·Computer Science
Develop a reinforcement learning framework in which two agents select features and a third selects an operation, with a shared value network, to generate new features in a traceable way.
Vijay Subramanian
University of Michigan, MI·Engineering
Develop robust reinforcement learning algorithms that use model-class knowledge, future predictions and off-policy methods to cope with mismatches and limited data from simulators or offline records.
The school
William & Mary: 11 labs on PI Pages, $95M a year on research, 67 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.