From: mailbox@dmitry-kazakov.de   
      
   On 3 Jun 2006 04:47:35 -0700, Hadi Asghari wrote:   
      
   > Thanks for your answer. I have actually studied the problem domain for   
   > some time and have a relatively good understanding of it, but my   
   > problems are related to applying that knowledge in fuzzy. Finding the   
   > inputs and writing the rules is easy, but I have no clue on optimizing   
   > the membership functions and on choosing the correct activation   
   > methods, etc.   
      
   In my view, the activation function, and/or operators and defuzzification   
   are determined by the problem space. When you take PROD and SUM you   
   silently assume that the truth values behave as probabilities of   
   independent events. Only the problem space may tell if they indeed are.   
      
   > I am using a trial and error approach, so more   
   > specifically, is there some 'general recommendation' in ranking   
   > problems?   
      
   IMO ranking could be mapped to truth values. I would try to formulate it as   
   u(A|x) < u(B|x), where u is the measure you took (in your case it is   
   probability, because you have chosen PROD and SUM). A and B are things you   
   compare, x is the condition (the observed state and actions you plant to   
   undertake.) An alternative approach is. If there is some natural ranking in   
   your domain, though difficult to evaluate, then you could define linguistic   
   variables on that rank. Comparing sets over linguistic variables should   
   again give you truth values.   
      
   However, as you said, it works rather unsatisfactory, so I'd assume that   
   the probabilistic model is inadequate in your case. This is why I believe   
   you should step back and learn more about the domain. You should adequately   
   model it, to make things working.   
      
   --   
   Regards,   
   Dmitry A. Kazakov   
   http://www.dmitry-kazakov.de   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)