From: rshepard@salmo.appl-ecosys.com   
      
   In article <4efc665.0309070434.4d255238@posting.google.com>, Andrzej Pownuk   
   wrote:   
      
   > My personal opinion is the following:   
      
   > 1) If you deal with uncertainty in natural language you should apply   
   > something like fuzzy set theory (however I have a lot of remarks to that,   
   > but this is only my problem :)) For example: If you see some object and   
   > you don't know whether it is a car or bicycle or motorcycle then you can   
   > describe this kind of uncertainty by using fuzzy set theory (well, you can   
   > try ...).   
      
     Perhaps your confusion arises from your mental models of the problems.   
   When considering physical models of objects (for example, the bicycle and   
   motorcycle mentioned above) if you are uncertain into which category to   
   place the object then it may be a vision problem, or the fact that the   
   object is a hybrid. The uncertainty is not related to natural language.   
      
     However, if you want to describe the bicycle or motocycle in subjective   
   terms such as "large", "fast", "cool" and so on, then you are legitimately   
   in the realm of fuzzy sets and fuzzy logic. That is because there is   
   inherent uncertainty in how big is a "large" object, at what speed it can be   
   considered "fast", or at what age you are to see it as "cool".   
      
   > 2) If you measure something and each measurement give you slightly   
   > different results, then you can describe this problem by using probability   
   > theory.   
      
     That's true. And you cannot measure "tall", "heavy", "pretty" or other   
   abstract concepts.   
      
   > However if you measure for example bicycle, then in each measurement you   
   > have to have some bicycle.   
      
     If you are measuring a bicycle, then all measurements represent some   
   physical aspect of that bicycle. Not _some_ bicycle, but _the_ bicycle.   
      
   > 3) Sometimes you don't know the exact value of some parameters but you   
   > know the range. In such cases you have set valued uncertainty.  If you mix   
   > this uncertainty and probability, then you can get for example random set   
   > and Dempster-Shafer's model (or other imprecise probability models).   
      
     You seem a bit uncertain. Some concepts -- such as "heat" -- can be measured   
   to arbitrary accuracy and precision using both absolute and ratio scales. We   
   all accept that the difference in heat content between 5-10 C is the same as   
   the heat content difference between 50-55 C. However, the concept of   
   "temperature", while often used as a synonym for "heat", is inherently   
   uncertain, imprecise, vague ... or fuzzy. Many of us would consider the   
   "temperature" difference between 5-10 C as "large", "significant" or "nice"   
   while that between 50-55 C could be considered "negligible", "still too hot"   
   and so on.   
      
   Rich   
      
   Dr. Richard B. Shepard, President   
      
                          Applied Ecosystem Services, Inc. (TM)   
               2404 SW 22nd Street | Troutdale, OR 97060-1247 | U.S.A.   
    + 1 503-667-4517 (voice) | + 1 503-667-8863 (fax) | rshepard   
   appl-ecosys.com   
                            http://www.appl-ecosys.com/   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)