| 作成者 |
|
|
|
|
|
| 本文言語 |
|
| 出版者 |
|
|
|
|
|
| 発行日 |
|
| 開始ページ |
|
| 終了ページ |
|
| 会議情報 |
|
| 出版タイプ |
|
| アクセス権 |
|
| 概要 |
This paper proposes two methods that improve the interface of inputing fitness values for interactive GA and evaluates them. The proposed are "input method of discrete fitness values (method #1)" and ..."combinational input method of discrete and continuous fitness values (method #2)." They let operators not care about distinguishing and rating slight differences of the interactive GA outputs. First, we evaluate if the quantized noise of the proposed method causes bad convergence of GA search. Simulation experiment has shown that the proposed methods do not result bad convergence in practical order. Then, we evaluate how the proposed methods reduce human burden through subjective tests. The statistical test has shown significantly better result than the conventional input method of continuous fitness values. It has also shown that the method #2 has better performance than the method #1.続きを見る
|
Mendeley出力
DomesticConf050