| authors |
Mackenzie, C.A. and Gero, John S. |
| year |
1986 |
| title |
Learning in the Domain of Decisions and Performances |
| source |
IAAI'86 Conference. 1986. pp. i:1:1-9. CADLINE has abstract only |
| summary |
Many domains present themselves as mappings between two classes of spaces: decision spaces and performance spaces. All design domains can be represented in this manner where the designer takes decisions which manifest themselves as performances in the designed artifact. Learning in these domains can take account of the structural characteristics of the spaces and of the mappings. This paper describes a system, PARE, which learns in the domain of decisions and performances by making use of the characteristics of a particular structuring concept known as 'Pareto optimality.' Much is known about the concept and its features which are used as hypotheses. If the hypotheses succeed then learning takes place by specializing the hypotheses' characteristics. Characterizations of Pareto optimality are described and the feature extraction process shown. The feature extraction process utilizes fuzzy pattern matching. An example of the system, written in ConSUN workstations, is presented from the domain of fenestration design |
| keywords |
performance, learning, design process, optimization, analysis, applications, theory, systems |
| series |
CADline |
| email |
|
| references |
Content-type: text/plain
|
| last changed |
2003/06/02 13:58 |
|
