dcbrown@wpi.WPI.EDU (David C. Brown) (05/22/91)
I have been teaching a graduate Expert Systems seminar for about 6 years based on the ideas I presented in the AI Magazine Vol.8 No.3, Fall 1987. The course needs updating. This is a request for advice. The course is organized around 3 hour presentations by small groups of students. Each group presents one expert system, based on reading a provided folder of papers from the literature. I chose the systems to be covered in order to provide a wide coverage across several dimensions -- for example, the task being done (eg. diagnosis, design, ...), the system architecture, reasoning techniques, and knowledge used (eg. rules, blackboards, generate & test, ...). I'll refer you to the AI Mag paper for more details. The students have already had a grad AI intro course. Each system demonstrates a different collection of ideas, and can be compared and contrasted against the others. The 12 systems chosen were: MDX [CSRL] MYCIN [EMYCIN] TEIRESIAS INTERNIST [CADUCEUS] CASNET-glaucoma PROSPECTOR AM [EURISKO] MOLGEN DENDRAL [Meta-DENDRAL] SU-X/HASP/SIAP R1/XCON AIR-CYL [DSPL] I know that this badly represents recent advances, such as case-based systems, systems that learn, model-based systems, and systems using qualitative reasoning. Given the goals and time constraints of the course, what expert systems should be replaced, by which other system, and why? Remember that replacements should be well-described in the literature. Thanks in advance for your suggestions, Dave Brown. ----------------------------------------------------------------------------- Dr. D. C. Brown, AI Research Group, DCB@CS.WPI.EDU Phone: (508) 831-5618 CS Dept., WPI, Worcester, MA 01609. FAX: (508) 831-5776