集合行為の効率性と公平性.pptVIP

07/10/00 The Design of Desired Collectives with Multi-Agent Simulation Akira Namatame Dept. of Computer Science National Defense Academy, Japan nama@nda.ac.jp Collectives of Interacting Agents Agent’s Behavior Based on the Logic of Minority Highlights of The Talk The inverse and forward problems Characterize the inverse and forward problems to self-organize desired collectives The interactive design with an agent-based model Propose the interactive design with multi-agent simulation. Logic of Minority: Asymmetric Problems Reasons for Undesirable Outcomes Most common observation (1) Bounded rationality of agents (2) Inconsistency between individual rationality and group rationality Points of this talk (1) Agents behave with false rules How do agents learn desirable rules? (2) Agents behave with inappropriate utility functions. How do agent should modify their own utility functions? Symmetric Problem vs. Asymmetric Problem Decomposition to Pair-wise Problems Desirable Collective: Stability, Efficiency, Fairness Characterization of Learning Models (1) Learning models without coupling with others Reinforcement learning Agents reinforce the strategy which gains the payoff Evolutionary learning Agents evolve strategy of interaction (2) Learning models with coupling Best-response learning Agents adapt based on the best-response strategy Coupling of Agents Coupling Rule between Two The Performances of Evolutional Learning What Agents Acquired with Evolutinary Learning ? 400 agents with different rules at the beginning evolved to share one of 15 coupling rules. Commonality of Acquired Rules The 15 meta-rules shared by all agents have the commonality Coupling with Local Neighbors Simulation Results Efficient and equitable dynamic orders are emerged with give-and-take Coupling Agents with Collectives (1)The action variable of agent Ai , a1(t) = 1 : S1 (Go) a1(t) = 0 : S2 (Stay) (2)The Status