Nie jesteś zalogowany | Zaloguj się
Facebook
LinkedIn

Finding Subgame Perfect Extreme Equilibria

Prelegent(ci)
Krzysztof Żyndul
Afiliacja
MIMUW
Język referatu
polski
Termin
14 kwietnia 2026 10:15
Pokój
p. 4060
Tytuł w języku angielskim
Finding Subgame Perfect Extreme Equilibria
Seminarium
Seminarium "DeSeR: Dane, strumienie, rozpraszanie"

When modeling multi-agent systems using simple stochastic games on graphs, standard Nash equilibria ignore how differently players perceive risk. The recently introduced Extreme Risk-Sensitive Equilibria (XRSE) addresses this through the concepts of optimists and pessimists: players who perceive their reward as either the maximum or minimum payoff achievable with non-zero probability. However, XRSE inherits a flaw from traditional Nash equilibria: it allows for irrational behaviors, such as non-credible threats. We refine this model and study Subgame Perfect Extreme (Risk-Sensitive) Equilibria (SPXE). In this talk, I will unpack these core concepts and show the computational complexity lower bounds and memory requirements for SPXE.

Joint work with Léonard Brice, Thomas A. Henzinger, and K. S. Thejaswini.