From: mailbox@dmitry-kazakov.de   
      
   On 24 Apr 2004 13:16:25 -0700, predictr@bellatlantic.net (Will   
   Dwinnell) wrote:   
      
   >Earl Cox wrote:   
   >"All of this stems, often subconsciously, from a sense that Boolean   
   >logic is the only CORRECT form of logic and that all other logics are   
   >inferior. This is, indeed, the gist of your argument."   
   >   
   >Dmitry A. Kazakov  responded:   
   >"Again note the difference between the models. If a problem is   
   >formulated in terms of Boolean logic, then clearly it is a correct   
   >form to deal with that. If a problem is formulated in fuzzy terms,   
   >then Boolean logic becomes inapplicable, if not extended   
   >appropriately."   
   >   
   >Isn't it the solution that is "formulated", not the  problem?   
   >Problems simply are.   
      
   Well, I always try to avoid philosophical questions related to the   
   question what is the "real world". To me a problem begins to exist   
   only when it is formally written.   
      
   You refer to "real world" problems, which can be formalized/formulated   
   in many different ways, from Boolean logic to fuzzy logic.   
      
   >While some problems naturally lend themselves   
   >to one type of solution over others, I think it is the analyst's   
   >decision to select an appropriate solution for the problem.   
      
   Of course.   
      
   >I do understand your contention that analytical solutions which   
   >proceed from domain-specific theory are preferable,   
      
   How could it be otherwise? I can evaluate Pi using Monte Carlo. I   
   could even try to do it with fuzzy using a system querying human   
   experts. (:-)) But why should I, if there is a deterministic algorithm   
   which works?   
      
   >and I suspect that   
   >your work lies in a field in which such solutions are available, but   
   >not all problems have such ready answers.   
   >   
   >I do not understand your contention that one would establish some   
   >arbitrary ordering of preference of solutions before ever seeing the   
   >problem.   
      
   It is not me, it was Occam, who brought that idea. (:-))   
      
   Consider a *new* "real world" problem. It is clear that better to try   
   to formulate it in terms of Boolean logic first. If that does not   
   work, then one should consider a statistical appoach, which is the   
   second choice. Third would be to try fuzzy and mixed approaches.   
   Disagree?   
      
   >Further, technical problems are not theoretical   
   >abstractions: they occur within a context which may impose   
   >requirements or preferences other than technical accuracy, such as   
   >transparency to non-technical people (like regulators) and ease of   
   >maintenance.   
      
   Add here customer preferences. Some of them do not want to even hear   
   about a system giving answers in terms of probabilities (it should   
   tell "yes" or "no".) Some of them require a NN-based system regardless   
   the problem. Most of them require a definite programming language, etc   
   (all that are real stories.) 70% of cases these requirements are   
   rather arbitrary and based on a current hype. Should it mean that the   
   answer to "when not to apply fuzzy" reads "when your customer tells   
   so"?   
      
   --   
   Regards,   
   Dmitry Kazakov   
   www.dmitry-kazakov.de   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)