Bridger: Learning
Bridge Design Knowledge
Bridger is an experimental learning system developed to explore the
possibility of using machine learning techniques for the acquisition and
improvement of design knowledge. Bridger is a multistrategy
learning system consisting of two main learning systems: ECOBWEB and EPROTOS. Both systems
employ multistrategy learning and both are case-based
reasoning systems. Thus Bridger is a multistrategy
learning system at the micro and macro sense. As shown bellow,
Bridger's architecture also integrates finite element analysis for checking a
newly synthesized bridge according to AASHTO requirements.
Bridger's architecture
A
Sample Publications
Include:
- Reich, Y. (1991), Design
Knowledge Acquisition: Task Analysis and a Partial Implementation, Knowledge
Acquisition, 3(3):237-254.
Contrary to views presented before 1990, this paper shows that by breaking
up design tasks into smaller subtasks, computational support can be
designed to aid the acquisiton of knowledge for
a complete design task. (Postscript
file, 273K, one figure missing)
- Reich, Y. and Fenves, S.
J., (1992), Inductive Learning of Synthesis Knowledge, International
Journal of Expert Systems: Research and Applications, 5(4):275-297.
This paper presents ECOBWEB, a system that extends the conceptual
clustering system COBWEB. Part of the paper shows how ECOBWEB traces a
change in a domain as reflected by bridges constructed in Pittsburgh. (Postscript file,
451K)
- Reich. Y. (1993), A Model of Aesthetic Judgment in Design, Artificial
Intelligence in Engineering, 8(2):141-153.
This paper presents a model of how aesthetic judgment in design may be
performed. It illustrates the ideas by using examples of bridge shaped
synthesized by Bridger. (Postscript
file, 810K)
- Reich, Y. (1993), The Development of Bridger: A Methodological
Study of Research on Machine Learning in Design, Artificial
Intelligence in Engineering, 8(3):217-231.
This paper presents a reflection on the course of the development of
Bridger. The aim is to give a more accurate description of the steps that
were carried out and their underlying reasons than is usually given in a
paper. (Postscript
file, 359K)
- Reich, Y. and Fenves, S. J.
(July, 1995), A system that Learns to
Design Cable-Stayed Bridges, Journal of Structural Engineering,
ASCE, 121(7):1090-1100.
This paper presents the most complete description of Bridger. (Note
however that due to ASCE's limit on the number
of words per paper [10,000] there are still details that were omitted.) (Postscript file,
408K; Zipped PDF
file, 855K)
- Reich, Y. (1995), The Value of Knowledge, International
Journal of Human-Computer Studies, 42(1):3-30.
The issue of knowledge evaluation is raised often in relation to building
intelligent systems, knowledge acquisition, or machine learning. There are
many proposals, both quantitative and qualitative that float around. This
paper organizes these by using the ideas of measurement theory. The paper
illustrates the ideas through the evaluation of the design knowledge
acquired by Bridger,
an experimental system for learning knowledge on cable-stayed bridge
design. (Postscript
file, 136K compressed; PDF,
185K zipped)
Related topics:
Copyright ©
1995-2005 Yoram Reich
Page URL: http://www.eng.tau.ac.il/~yoram/bridger.html
Last
modified: 4/18/2005 9:10:00 PM