barash@mmlai.UUCP (Rev. Steven C. Barash) (02/09/88)
Does anyone reading this understand "Fuzzy set theory"/"Fuzzy logic"
and its applicability to automated reasoning?
In particular, I'm interested in how one might verify empirically
(or experimentally, as with probability theory) the accuracy of the
fuzzy set formaulas for appropriate domains. Also, for a given problem,
how should one determine the suitability of fuzzy sets (instead of traditional
methods) for reasoning under uncertainty? The journal articles
tend to be rather specialized, and don't address such basic issues.
Please repond by E-mail; I'll post a summary if interest is sufficient.
Any ideas will help, and thanks in advance.
Steve Barash
--
Rev. Steve Barash @ Martin Marietta Labs / Artificial Intelligence Department
Disclaimer: I speak for no one.
ARPA: barash@mmlai.uu.net
UUCP: {uunet, super, hopkins!jhunix} !mmlai!barashvu0112@bingvaxu.cc.binghamton.edu (vu0112) (02/10/88)
In article <275@mmlai.UUCP> barash@mmlai.UUCP (Rev. Steven C. Barash) writes: > >Does anyone reading this understand "Fuzzy set theory"/"Fuzzy logic" >and its applicability to automated reasoning? I'm trying to. . . >In particular, I'm interested in how one might verify empirically >(or experimentally, as with probability theory) the accuracy of the >fuzzy set formaulas for appropriate domains. I'm not sure how such verification would differ from that for crisp formulas. >Also, for a given problem, >how should one determine the suitability of fuzzy sets (instead of traditional >methods) for reasoning under uncertainty? First, obviously, if the system in question is non-deterministic, then fuzzy methods must come into play. It should be recognized that probability theory is a special case of fuzzy theory. Now, as to the question of whether to use non-probabilistic (e.g. possibilistic) fuzzy methods, that depends on the law of the excluded middle (True(A) => False(~A)), which probability conforms to, and possibility does not. If the samples are highly interdependant, fuzzy can yield better results. I recently wrote a paper on Fuzzy Artificial Inference and Expert Systems. Fuzzy promises a much more succesful, general method for approximate reasoning. >The journal articles >tend to be rather specialized, and don't address such basic issues. Try _Fuzzzy_Sets_and_Systems_. Also, I'd reccommend _Fuzzy_Sets, Ucertainty,_and_Information_ (George Klir, Prentic Hall 1988), which is an excellent introduction and bibliography. Read anything by Zadeh. >Please repond by E-mail; I'll post a summary if interest is sufficient. Sorry, couldn't resist. Plus my mailer usually chokes these days. > Steve Barash O----------------------------------------------------------------------> | Cliff Joslyn, Mad Cybernetician | Systems Science Department, SUNY Binghamton, Binghamton, NY | vu0112@bingvaxu.cc.binghamton.edu V All the world is biscuit shaped. . .