Co-evolution of Sustainable Coupling Rules

スポンサーリンク

概要

• 論文の詳細を見る

• In this paper, we show how self-interested agents collectively learn "correlated rules" that sustain both efficient and equitable dynamic orders. The conventional zero-sum RSP game is generalized as non zero-sum games by modifying payoff matrix. We formulate a collection of agents in a lattice model repeatedly play the generalized rock-scissor-paper (RSP) games. Agents are modeled to play with neighbors by applying interaction rules. Those interaction rules are evolved by crossing over with the most successful neighbor. If the payoff for the winning the game increases, agents learn to win and lose in a coordinated way: they repeat these coordinated plays in order to realize the most efficient and equitable situation. We also investigate learning rules of agents. We show interacting agents co-evolve to learn sharing the same types of rule, which sustain the efficient and equitable society.

• 一般社団法人情報処理学会の論文
• 2004-08-04

著者

• Namatame A

  Dept. Of Computer Science National Defense Academy Yokosuka

• Namatame Akira

  Dept. Of Computer Science National Defense Academy

• Sato N

  Dept. Of Computer Science National Defense Academy Yokosuka

• Sato Naoto

  Dept. of Computer Science, National Defense Academy, Yokosuka

• Sato Naoto

  Dept. Biol. Sci. Grad. Sch. Sci. Univ. Tokyo

• 佐藤 直人

  Dept. of Computer Science, National Defense Academy, Yokosuka

• 生天目 章

  Dept. of Computer Science, National Defense Academy, Yokosuka

関連論文

• Knowledge Sharing in an Organization of Heterogeneous Agents
• Knowledge Sharing in an Organization of Heterogeneous Agents (知能と複雑系研究報告)
• Co-evolution of Sustainable Coupling Rules
• ISOLATION AND CHARACTERIZATION OF TWO TOBACCO HOMOLOGS OF tb1, AN APICAL DOMINANCE-RELATED GENE IN MAIZE.

スポンサーリンク