Genetic Evolution of Communication in Distributed Classifier Systems

スポンサーリンク

概要

• 論文の詳細を見る

• Machine learning paradigms requiring a supervisor are not appropriate when faced with learning tasks for which no such supervision may be provided. For example, autonomous agents,such as robots cannot be given explicit hiformation about theadaptive behavior corresponding to different situations they mayencounter. Required knowledge niust be inferred through interaction with the environment. Furthermore, since this icnowlednot available ail at once, they must Iearn it incrementaily. TCesieStwo characteristics are the two most iniportant features of a classof learning methods called incrementai reinforcement learning.Classifier systems, proposed by Jolin Holland, are promisinglearning paradigin to fulfill these needs. They have been usedto study the behavior of adaptive organisms and autonomousagents.

• 一般社団法人情報処理学会の論文
• 1992-09-28

著者

• Okamoto Takuya

  University Of Tokushima

• Ono Norihiko

  University Of Tokushima

• Rahamani Adel

  University of Tokushima

関連論文

• Genetic Evolution of Communication in Distributed Classifier Systems

スポンサーリンク