AI-woordenlijst
Het complete woordenboek van kunstmatige intelligentie
Nash Equilibrium
Fundamental concept in game theory where no agent can improve their payoff by unilaterally changing their strategy, given that the strategies of other agents remain fixed.
Zero-Sum Game
Type of game where the total payoff of agents is constant, meaning that one agent's gain corresponds exactly to the losses of the other agents.
Equilibrium Point
Stable strategic configuration where no agent has an incentive to unilaterally deviate, representing a stationary state in multi-agent systems.
Best Response
Optimal strategy of an agent given the strategies of other agents, maximizing their expected payoff in the current game context.
Fictitious Play
Iterative algorithm where each agent assumes that opponents will use the empirical distribution of their past actions to determine their best response.
Multi-Agent Q-learning
Extension of the Q-learning algorithm to multi-agent environments where the Q-function depends on the joint actions of all agents.
Iterated Prisoner's Dilemma
Repeated game where cooperation emerges as an evolutionarily stable strategy despite the dominance of defection in the single-shot game.
Coordination Game
Category of games where agents benefit from choosing the same strategy, creating multiple equilibria that may be potentially suboptimal.
Decentralized Reinforcement Learning
Paradigm where each agent learns independently without central communication, based on its local observations of the environment.
Counterfactual Regret Minimization
Learning algorithm that minimizes counterfactual regret to converge to Nash equilibrium in extensive-form games with incomplete information.
Fundamental Minmax Theorem
Mathematical principle establishing the existence of Nash equilibrium in mixed strategies for finite zero-sum games.
Correlated Equilibrium
Equilibrium concept where an external signal correlates agents' actions without being binding, potentially improving collective efficiency.
Multi-Agent Policy Learning
Direct policy optimization approach in multi-agent systems, accounting for non-stationarity induced by learning agents.
Evolutionary Game Theory
Theoretical framework applying natural selection principles to the dynamic analysis of strategies in populations of agents.
Stochastic Games
Extension of Markov decision processes to multi-agent environments where transitions and rewards depend on joint actions.
Convergence to Equilibrium
Property of learning algorithms guaranteeing asymptotic approach to an equilibrium point under certain game conditions.