標題: 空間配置規則之推導方法－人工智慧中案例學習法之運用The Induction of Rules for Spatial Layout - The Utilization of Learning by Examples in AI 作者: 徐開平Hsu, Kai-Ping黎漢林Li, Han-Lin資訊管理研究所 關鍵字: 空間學習;機器學習;案例式學習;空間配置問題;空間語法;spatial learning;machine learning;learning by examples; spatial layout problem;spatial grammar 公開日期: 1994 摘要: 本文運用人工智慧中案例學習法，來推導空間配置之規則。首先針對複雜 度不同的配置個案，設計出兩種物件與空間關係的語法，再以此導出一空 間配置案例之學習方法。文中並分別以廚房配置及住宅平面設計為例，進 行配置規則的推導。本方法相對於ID3 法及D-MAX 法之優點為：1.ID3 法 推導出的規則相當繁鎖，且普及率較低。而本方法推導出的規則較精簡且 正確率及普及率均較高。2.D-MAX 法所需的計算時間太久，而本方法的計 算時間相當短，並不受限於配置案例及相關屬性的多寡，均適用。 This thesis uses the method of learning by examples in artificial intelligent to induce the rules of spatial layout. First, two kinds of grammar about the relationship of objects and spaces for two examples of different complexity are designed. Following that, a learning method for inducing the rules of spatial layout is proposed. The examples of the kitchen layout and the house design are used to induce the rules of layout. Comparing the proposed method with the ID3 and the D-MAX shows that: 1.The rules induced by the ID3 are not compact and the coverage rate of the rule is low; While the rules induced by the proposed method are more precise and both of the accuracy and coverage rate are higher. 2.The computation- al time of D-MAX is quite long. Nevertheless, the proposed method can obtain the result in a short time. URI: http://140.113.39.130/cdrfb3/record/nctu/#NT830396026http://hdl.handle.net/11536/59133 Appears in Collections: Thesis