Get Started
Home
Topics
Search
Library
Research questionHow can we learn welfare-optimal approximate Nash equilibria in concurrent stochastic games under uncertain transitions while certifying exact-equilibrium nonexistence?Limited trajectory data leaves the game’s transition dynamics uncertain, making equilibrium computation unreliable. The learning procedure must also distinguish an approximate equilibrium from cases where no exact Nash equilibrium exists.
Evaluation & Benchmarks
Machine Learning
Multi-agent Systems
Reinforcement Learning
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.Robust PAC Learning of Concurrent Stochastic GamesThe source studies finite-horizon concurrent stochastic games with uncertain transition kernels, using data-driven L1 confidence sets and a minimum reachability condition for relevant state-action pairs. It reports polynomial trajectory-sample guarantees and benchmark experiments, but the evidence is limited to the stated game setting and assumptions.research paper · Sep 3, 2026
Related questions
How can sequential allocation of indivisible goods preserve EF1 while retaining high Nash social welfare under identical additive valuations?How can we learn near-optimal policies from transition samples in large or infinite-state-action MDPs using function approximation?How can uncoupled learning keep individual regret bounded independently of time in arbitrary multi-player normal-form games?How can imperfect-information game solvers build useful state abstractions without domain-specific evaluators?