Newsgroups: comp.ai.fuzzy
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!europa.eng.gtefsd.com!howland.reston.ans.net!agate!ames!news.hawaii.edu!uhunix3.uhcc.Hawaii.Edu!pollarda
From: pollarda@uhunix3.uhcc.Hawaii.Edu (Art Pollard)
Subject: Fuzzy as a NN replacement?
Message-ID: <D2GpC1.Bw0@news.hawaii.edu>
Sender: news@news.hawaii.edu
Organization: University of Hawaii
Date: Sun, 15 Jan 1995 19:40:49 GMT
Lines: 21



I have a pattern recognition algorithm (for characters in graphics i.e., 
OCR) that uses Neural Nets to make the final decision.  The algorithm uses a 
histogram which was created from the graphical mark which was extracted 
from the page.  

The problem as I see it with Neural Nets is that you can't get 
probabilities and so, you only get one answer.  I am new to the fuzzy 
scene but, was thinking that there might be some sort of fuzzy algorithm 
where I could give it a histogram of a character and it would tell me 
which characters it could be and what would be the probability of each.
I could then use the answer(s) for further processing.

If fuzzy isn't the right choice for this, is there some other algorithm 
which might work?

Thanks for any help you can provide,

Art

