YZ Sözlüğü
Yapay Zekanın tam sözlüğü
Multi-Agent Reinforcement Learning
Learning paradigm where multiple agents simultaneously learn to make decisions in a shared environment, interacting with each other to optimize collective or individual objectives.
Multi-Agent Deep Deterministic Policy Gradient (MADDPG)
CTDE algorithm extending DDPG to multi-agent environments, using centralized critics and decentralized actors to learn in continuous action spaces.
Multi-Agent Partially Observable Markov Decision Process (MPOMDP)
Mathematical formalization of MARL environments where each agent has partial observations and must infer the global state to make optimal decisions.
Mean Field Games
Theory studying the interactions of a large number of rational agents by approximating the crowd effect through a mean field, applicable to large-scale multi-agent systems.
Continuous Control
Application domain of MARL where agents must control physical systems with continuous actions, such as mobile robotics or object manipulation.
Stochastic Games
Extension of MDPs to multi-agent environments where transitions and rewards depend on the joint actions of all agents, modeling cooperative and competitive scenarios.
Nash Equilibrium in MARL
Stability concept where no agent can improve its reward by unilaterally changing its strategy, used as a convergence criterion in competitive MARL algorithms.
Coordination Protocols
Communication or synchronization mechanisms allowing agents to align their actions to achieve collective objectives in continuous MARL environments.