From: ferry@onthenet.com   
      
   On 1 Apr 2004 08:54:04 -0800, wsiler@aol.com (William Siler) wrote:   
      
   >TheFerryman  wrote in message news:...   
   >> When using the Combs method should I expect to get exactly the same   
   >> inferred result as I get from my standard rule base? At the moment the   
   >> results differ slightly. Not enough to make any difference to the   
   >> performance but I was under the impression they should be identical.   
   >   
   >Indeed, you can expect slightly different results, even with the same   
   >expert system shell. However, this is to be expected since the Combs   
   >method is very different from the conventional method. That does not   
   >mean that the conventional method is RIGHT; there are at least three   
   >methods of solving the Fisher Iris problem. The conventional method,   
   >of assigning membership functions to descriptors such as Small, Medium   
   >and Large works poorly. Assigning membership functions to the   
   >different attributes for different species (setosa, virginica,   
   >versicolor) works very well, as does the Combs method with these   
   >membership functions.   
   >   
   >William Siler   
      
   Thanks for your answer William. I have no idea what the Fisher Iris   
   problem is but I'm sure it must be relevant ;-)   
      
   Here's the follow up question then... I have read Combs' paper and   
   followed his proof, which to an uneducated man like myself makes   
   perfect sense. However *if* the proof is correct and a rule base is   
   converted between systems correctly how is it possible that the   
   results differ?   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)