建築設計推理中精確性與粗略性的初探 - 設計運算中符號論與連接論的研究

Abstract

透過假設的說明，本研究將有兩個目的：主要目的在討論於建築設計推理 的精確性與粗略性，以符號理論與連接理論作為兩者的模擬基礎，各自建 立其系統的知識表達方式。另一個目的在於模擬精確性與粗略性不同知識 表達的傳遞方式。方法上以知識基礎系統系統(knowledge-based systems KBS)與神經網路(neural networks)模擬建築設計推理中的精確性( exactness)與粗略性(scruffiness)的步驟。其中知識基礎系統的部分以 迪化街的店舖住宅平面的設計知識為知識庫的基礎；神經網路採用逆傳遞 模型(back propagation BP)建立，以國外的建築平面案例為神經網路的 樣本。步驟上的第一階段我們將把迪化街的店舖住宅為樣本建立系統，把 這個系統視為一種精確性的推理方式。當設計推導無法符合其條件的情況 下，我們假設此時會發生粗略性的推理方式。第二階段是建立神經網路系 統與兩個知識表達的傳遞方法，前者是我們假設粗略性推理方式的部分， 以國外的建築案例為樣本；最後，討論的是當我們建立推理方式精確性與 粗略性的部分之後，如何兩者之間如何傳遞訊息，本研究是將粗略性推理 機制的訊息傳遞到精確性的推理機制，然後「調適」(adaptation)為精確 性推理機制的知識表達。對於建築的設計推理研究，我們提供了傳統在精 確推理機制下另一個屬於粗略性的初步探討，除此之外，也涉及了人工智 慧中兩種不同研究取向的初步結合，或許可以作為電腦輔助建築設計( CAAD)的基礎研究結果。 There are two main purpose of this research: first, study the exactness and scruffiness in architectural design reasoning by symbolism and connectionism theory; second, imitate the different knowledge representation of the exact and scruffy reasoning.We use the knowledge-based systems and neural networks to simulate the steps of exactness and scruffiness of design reasoning in architectural. The KBS is based on the research of the shop house in Di-Hwau Street, and the neural network is established by back propagation, and based on the plan drawing of architecture masterpiece.The first step of this research is taken the shop house of Di-Hwau Street as sample, and taken this sample as a reasoning of exactness. We assume this moment would process the scruffiness reasoning while the design refinement could work.The second step of this research is establish neural network and two style of knowledge representation, the former is the hypothesis of the scruffiness reasoning, and take the architecture masterpiece as sample; Finally, disscused the interaction between the exactness and scruffiness we established. This research is trying to put the scruffiness reasoning into the exactness, and adapt as the the knowledge representation of the exact reasoning.This research is trying to offered the scruffiness reasoning into the exactness one, besides, it evolve the combination of two different research, could be the foundation of the CAAD research