bharat@gaea.uucp (R. Bharat Rao) (01/02/90)
E-mail to : bharat@gaea.cs.uiuc.edu As part of my Ph.D. thesis I need to do empirical discovery, and discover relationships between groups of variables. These variables may be real-valued, nominal, discrete (different domains), or combinations of these types, and varying amounts of background knowledge (often none) will be provided. The data also will be (known to be) noisy in certain domains. I am looking for programs that can do empirical discovery - as many as I can find - and I wish to apply them to domains the programs are suited to. If you have the source code, or pointers to the code + literature on discovery programs that I could use, I would be very grateful if you could e-mail me the details at the address above, or post a response to comp.ai. I will summarize and post the results. Thanks, -Bharat R.Bharat Rao E-Mail: bharat@gaea.cs.uiuc.edu US Mail:AI Group, Beckman Institute, Univ. of Illinois, Urbana-Champaign.
Nagle@cup.portal.com (John - Nagle) (01/05/90)
Presumably you are familiar with AM and Eurisko. A recent
article on this subject is Lenat's "Why AM and Eurisko Appear To
Work" in Huberman's "Computational Ecology", which is a little outside
the mainstream AI literature and worth checking out.
This isn't a field in which much work has been done recently.
There were a few easy hits in the early days, but it proved very
difficult to advance much beyond the level seen in AM/Eurisko.
But Lenat, who is at the MCC in Austin, is still trying.
John Nagle