Tuomas Sandholm
Carnegie Mellon University, PA·Computer Science
Many important strategic situations, such as negotiations, auctions, cybersecurity and defense, are far too large or complex for people to solve by hand. Often players cannot see everything their opponents know, and sometimes the rules are not even written down, so a solver can only practice inside a simulator.
- Fields
Computer Science
Mathematics & Statistics
- Topics
game theory
computational game solving
imperfect-information games
- Grants
- 1 known grant, through Jul 2027
What they work on
- Design, implement and test scalable subtree-solving techniques that refine strategies for the part of a multi-step, imperfect-information game actually reached during play.
- Develop subgame-solving methods that work when the common knowledge closure, the set of game states consistent with public knowledge, is too large to compute directly.
- Create scalable techniques that find equilibrium strategies with guarantees of low or zero exploitability when the solver has only a simulator and no written rules.
More about the research
This project aims to make computer algorithms that find good strategies for these settings far more scalable. It focuses on multi-step games of imperfect information, where players act in turns and hold hidden information.
Why it matters
Algorithms that compute strong strategies with guarantees could support decision-making wherever parties interact strategically. The abstract names negotiation, business, defense, auctions, strategic pricing, cybersecurity and medicine. Because the methods work from a simulator alone, they could also apply to settings where the rules are unknown, such as trading simulations and real-time strategy environments.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Simina Branzei
Purdue University, IN·Computer Science·Early-career
Analyze market dynamics in which computationally limited players with only local information interact over time, testing whether globally efficient allocations emerge and whether markets grow or shrink.
Asaf Cohen
University of Michigan, MI·Mathematics & Statistics
Rigorously connect the long-horizon behavior of finite-player stochastic games to their mean field game limits, identifying structural features that stay stable as the number of players grows.
Igor Cialenco
Illinois Institute of Technology, IL·Mathematics & Statistics
Develop a tractable top-down stochastic model of groundwater levels and use it to study how prices of tradeable groundwater rights form as a Nash equilibrium among competing water users.
The school
Carnegie Mellon University: 28 labs on PI Pages, $505M a year on research, 346 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.