Abstract

Research on machine learning in design has concentrated on the use and development of techniques that can solve simple well-defined problems. Invariably, this effort, while important at the early stages of the development of the field, cannot scale up to address real design problems since all existing techniques are based on simplifying assumptions that do not hold for real design. In particular, they do not address the dependence on context and multiple, often conflicting, interests that are constitutive of design. This paper analyzes the present situation and criticises a number of prevailing views. Subsequently, the paper offers an alternative approach whose goal is to advance the use of machine learning in design practice. The approach is partially integrated into a modeling system called n-dim. The use of machine learning in n-dim is presented and open research issues are outlined.

Author Keywords: machine learning; design; computational models; design practice; flexible modeling; shared memory; natural language processing; multistrategy learning

Article Outline

• References