大富豪におけるペア温存戦略基準の獲得

情報処理学会

書誌事項

タイトル別名
  • Acquisition of pair keeping strategy standard in DAIFUGOU game

説明

本論文では，不完全情報ゲームであるトランプゲームの大富豪を対象として，不確定な状況への適応学習について考察した．まず，予備実験により，ペア温存戦略が有効であることが確認できた．しかし，大富豪では，対戦相手・ルールによってペア温存戦略の基準が変わってくる．そこで，ペア温存戦略を動的に学習する手評価学習を実装した．手評価学習では，対戦結果に応じて，各手の評価値を更新する．対戦実験の結果，対戦相手・ルールに応じたペア温存戦略の基準が獲得できた．

In this thesis, the adjustment study of the uncertainty to the situation was considered for the DAIFUGOU game that was the imperfect information game. At first, it was able to be confirmed that the pair keeping strategy was effective by a preliminary experiment. However, the standard of the pair keeping strategy changes in the DAIFUGOU game according to the opponent players and the rule. Then, we implemented the play evaluation learning that dynamically studied the pair keeping strategy. In the play evaluation learning, the evaluation value of each play is updated according to the game result. As a result of the experiment, the standard of the pair keeping strategy corresponding to the opponent players and the rule was able to be acquired.

収録刊行物

詳細情報 詳細情報について

問題の指摘

ページトップへ