다층 퍼셉트론의 맥락으로 본 디자인 콘셉트 개발 모형

Abstract

대부분의 디자이너들은 디자인 콘셉트를 구상할 때 많은 시간과 노력을 요한다. 본 연구는 이런 불필요한 스트레스를 배제하고 디자인 콘셉트 구상을 보다 쉽게 하기 위해 디자인 콘셉트 개발 방법론을 제안한다.

연결주의의 네트워크 모형은 지속적인 발전을 거듭하여, 현재 30여 개의 모형이 존재하고 있다. 본 연구자는 기본적인 다층 퍼셉트론 모형을 응용한 결과, 다층 퍼셉트론적 콘셉트 개발이 가능하다는 것을 기초 연구를 통해 발견하였다. 그 다음 단계로 본 연구에서는 다층 퍼셉트론에 의한 콘셉트 개발에 적합한 모형을 개발하기 위해, 다양한 다층 퍼셉트론 모델 중에 가장 널리 영향을 끼친 전방향 네트워크 모델인 맥클런드(McClelland)와 러멜하트(Rumelhart)의 ‘단어 지각 모델’(Word Perception Model)을 적용한 다층 퍼셉트론에 의한 디자인 콘셉트 개발 모형을 만들었다. 개발된 모형은 디자인 문제가 주어지면 문제의 가치를 구분 짓고, 은닉 단위와 층 차를 연결하는 인과관계 종류를 계획하여 효율적으로 디자인 콘셉트를 개발하게 한다.

Many designers have spent many hours and efforts thinking of and developing design concepts. This study is about formulating a methodology for design concepts to eliminate the un-necessary stress and to plan a design concept efficiently.

The network models of connectionism have been experiencing growth continually. There are about 30 models existing today, and they are studied by many disciplines. Utilizing a basic Multilayer Perceptron model, the study found that concept development is possible utilizing the Multilayer Perceptron methodology. The initial study reconfirmed the possibilities, and then the final study initiated by utilizing the most important feedback network model used by many disciplines, 'Word Perception Model' of McClelland and Rumelhart, to find a model of the design concept methodology by Multilayer Perceptron. The result shows that it is efficient and easy to plan a design concept through defining values of a given design problem, and planning the types of cause & effect relationships for connecting hidden nodes and layers.;Many designers have spent many hours and efforts thinking of and developing design concepts. This study is about formulating a methodology for design concepts to eliminate the un-necessary stress and to plan a design concept efficiently. The network models of connectionism have been experiencing growth continually. There are about 30 models existing today, and they are studied by many disciplines. Utilizing a basic Multilayer Perceptron model, the study found that concept development is possible utilizing the Multilayer Perceptron methodology. The initial study reconfirmed the possibilities, and then the final study initiated by utilizing the most important feedback network model used by many disciplines, 'Word Perception Model' of McClelland and Rumelhart, to find a model of the design concept methodology by Multilayer Perceptron. The result shows that it is efficient and easy to plan a design concept through defining values of a given design problem, and planning the types of cause & effect relationships for connecting hidden nodes and layers.