2-406 マルチエージェント環境における強化学習の構成法に関する一考察 : 実例に基づくアプローチと個体識別能力の効用

書誌事項

タイトル別名
  • Design of Reinforcement Learning in a Multi-agent Environment : Instance-based Approach and the Effectiveness of Indivisual Recognition

この論文をさがす

抄録

An approach to developing social behavior in a multi-agent environment by applying instance-based reinforcement learning to each indivisual and the effectiveness of indivisual recognition is investigated. As a result of comparative experiments with or without an indivisual recognition and extraction of dominant attributes of input data, we confirmed the effectiveness of indivisual recognition. Moreover, we confirmed the robustness of instance-based approach to large-scale and changing dynamic problem by extraction of the practical search space of reinforcement learning process.

収録刊行物

参考文献 (7)*注記

もっと見る

詳細情報 詳細情報について

問題の指摘

ページトップへ