From: @   
      
   EarlCox wrote:   
      
   > Bill,   
   >   
   > I must have missed Bruno's input to this question. But he has, as always,   
   > something sensible to say. In this case I agree with him.   
    >   
      
   Thank you, Earl. I copied my posting at the end of this message, after   
   the end of the quotation of your text. My answer to "luko" was within   
   the framework of my own work, mostly real-time embedded systems in the   
   last 15 years.   
      
   I do not know what "luko" does and what his affiliation is, given   
   that he writes from libero.it, a free Italian web-based mail system   
   that anybody can have for free as long as they have a land line in   
   Italy, at least these were the rules in 1997. I was a bit surprised   
   for his question because in Italy fuzzy logic enjoys a higher status   
   than in North America. In a sense, it is better known there. There   
   is less or no need to justify using it in Italy. During my last   
   3 trips I found that fuzzy logic is taught formally in all major   
   engineering schools in Italy and that a lot of thesis are written   
   yearly, even with some applications also in the social sciences.   
   I even bought a fuzzy logic book in Italian, because, though   
   fluent in Italian, I cannot speak about any technical or financial   
   subject without switching to English. I had the ambition to learn.   
   During my browsing in various university bookstores, I found that   
   all of their books were about microprocessor based applications.   
   SGS-Thompson Microelectronics has produced several application notes.   
   In 2000 I attended a conference on quantum computing and I found that   
   several Italian scientists were thinking about quantum computing   
   applications of fuzzy logic, hardware quantum fuzzy processors.   
   Zadeh, Kosko, and yourself were available in English and the   
   students were buying with interest, even outside course work.   
      
   I agree with what you say:   
    >   
   > I am not proposing   
   > that fuzzy logic is a solution to all kinds of problems -- only that the   
   > comments in this thread have not focused on what fuzzy logic actually means   
   > and what fizzy logic actually does and how it actually works.   
   >   
   > Anyone who is attacking real world problems must have a wide spectrum of   
   > tools. Consultants and researchers who attempt to fit every problem into a   
   > specific methodological class are setting themselves up for continual   
   > disappointment. But it is also important to actually understand the   
   > underlying mechanics, ontology, epistemology, and philosophy behind your   
   > tools so that you can select the right tool for each problem. My complaint   
   > is simply that there is a massive and deep misunderstanding of what fuzzy   
   > logic is all about and how it works. This has kept it from being applied in   
   > areas where it can yield dramatic results.   
   >   
   > Enuf said.   
   > Earl   
   >   
   I see some analogy between fuzzy logic and real-time embedded systems.   
   I have taught continuing education courses covering both topics   
   separately and even some course at conferences covering fuzzy logic   
   in real-time embedded systems. It seems to me that the best way to   
   transfer knowledge is by making people  shadow you when you work   
   and then you shadow them when they try. However, within a course   
   you cannot go beyond anything trivial. In both areas it is not   
   possible to give  step-by-step cookbook solutions, at least   
   not beyond some very simplistic example. Rather than the modern   
   university, the correct vehicle would be the guild workshop   
   of the renaissance, where apprentice used to follow  the old   
   master  and slowly learned by osmosis. How else can you transmit   
   "ontology, epistemology, and philosophy behind your tools"?   
   I could say the same thing about neural networks, genetic   
   algorithms, and variations of cellular automata.   
      
   When we arrive to the kind of applications that you have done,   
   in my view, things are even more difficult to do right and   
   to teach. I have been involved in very few and I feel that   
   could always be done better. In some case, I would  like to   
   re-do the job all over again to do it better (difficult after   
   the end of a contract :-) ).   
      
   Have a great weekend!   
      
   Bruno   
      
      
   > "William Siler"  wrote in message   
   > news:49b9df3d.0404182011.7337a053@posting.google.com...   
   >   
   >>"Bruno Di Stefano, P.Eng. (@ Nuptek Systems Ltd.)"   
   >   
   >  wrote in message   
   > news:<40808D7F.5020504@sympatico.ca>...   
   >   
   >>>The best approach to problem solving is to have a wide toolbox, with   
   >>>as many tools as possible, at least IMHO   
   >>>   
   >>   
   >>Amen, brother! Also in my (not so humble, after 65 years of problem   
   >>solving) opinion.   
   >>   
   >>William Siler   
   >   
   >   
   >   
      
   //////////////////////////////////////////////////////////////////////   
   ////////// Here's the original posting by Bruno Di Stefano ///////////   
   //////////////////////////////////////////////////////////////////////   
      
   Luko wrote:   
      
    > Come back to the first question.   
    >   
    > Where and why not use fuzzy?   
    >   
    > I can use it in deterministic or stochastic cases, and then where do i   
    > can't?   
    >   
    > Luko   
    >   
    >   
      
      
   Hi:   
      
   In general, I do not use fuzzy logic whenever I have or I can derive   
   an equation or a set of equations expressing the relationship between   
   input and output variables. This does not mean that I actually implement   
   the system as an equation or a set of equations. In real-time embedded   
   systems where the requested response time is very difficult to achieve,   
   I prefer to replace the equation(s) with one or more look-up tables.   
      
   There is a caveat associated with what I wrote: the case in which small   
   errors in reading input variables result in major erroneous swings of   
   the output. In this case, normally, I first try to make sure I minimize   
   errors in reading input variables. Depending on the physical system,   
   there are various things I can do, i.e. time filtering, etc. In the case   
   in which this is difficult, I can still use a fuzzy system, as fuzzy   
   system can be designed to exhibit better error/fault tolerance. Maybe,   
   I should not say "better", but I should say that they are easier to   
   design with error/fault tolerance.   
      
   As a general issue, not related just to fuzzy systems, "When the only   
   tool you have is a hammer, everything looks like a nail".   
      
   I do not claim to have answered the question. There are other cases   
   when in which maybe I would not use fuzzy system. This is the most   
   obvious.   
      
   The best approach to problem solving is to have a wide toolbox, with   
   as many tools as possible, at least IMHO   
      
   Ciao   
      
   Bruno Di Stefano   
   fuzzy.logic.nuptek@sympatico.ca   
      
      
   [continued in next message]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)