From: earlcox@earlcoxreports.com   
      
   Well that's not totally true. In a fuzzy expert system (mine for instance)   
   you could   
   write something like,   
      
      if costs(t-1) are very high   
        and income(t) is somewhat low then do;   
            risk is elevated;   
            if reserves(t+1) are low then   
                cashflow is reduced;   
        end do;   
      
   the value of the outcome fuzzy set "risk" is correlated with the truth of   
   the two antecedent propositions. The second consequent if-then statement   
   uses this antecedent truth as a weighting or scaling factor for the truth of   
   its premise. Thus the outcome fuzzy set "cashflows" is correlated with the   
   truth of the nested rule statement as well as the truth of the controlling   
   rule statement (which makes sense, the nested rule cannot be truer than the   
   truth of the fuzzy proposition that caused the nested rule to be executed.   
   And the outcome of the nested rule cannot be truer than the truth of the   
   premise that contributed to its under-generation fuzzy set.)   
      
   At the same time, it is well to bear in mind that maintainability of a large   
   and complex fuzzy system is important. The ability to specify several   
   outcomes that are correlated by the same predicate almost always means that   
   the rule set is smaller and easier to understand, debug, and extend. This is   
   particularly true if the common predicate contains a number of propositions   
   with, perhaps, complex expressions or lead/lag time relationships. In a   
   financial system, as an example, a rise in the debit to equity or debt to   
   income ratio in a set of previous periods, compounded with a rise in the   
   projected cost of capital in the next quarter compounded by a reduction in   
   available inventories might produce several outcomes: a reduction in the   
   client's credit worthiness, an acceleration of a loan repayment schedule,   
   the postponement of investment opportunities, and so forth. It would be   
   extremely awkward not to mention difficult to maintain and understand if   
   each complex premise had to be repeated in the rule base for each of the   
   outcomes.   
      
   Assertions that multiple outcomes are simply an AND condition and can be   
   handled by repeating the rule arise from individuals who work only in small   
   control systems (such as academic exercises) or who have never actually   
   built and delivered operational fuzzy models for the business world.   
      
   Enough said.   
   Earl   
      
   "Bartosz Bien"  wrote in message   
   news:btc2tk$qhr$1@nemesis.news.tpi.pl...   
   > "tarmat"  wrote in message   
   > news:o7pivvoko10o0b4r3ok6pm67g91qebq097@4ax.com...   
   >   
   > > Can a rule have more than one consequence? If so, are the values   
   > > defuzzified in the usual way? Do you have any references to systems   
   > > that make use of multiple consequents?   
   >   
   > As the multiple consequents would be connected with AND operator, such a   
   > rule would easily be splittable in two or more equivalent rules. This is   
   why   
   > they are not used (at least I haven't seen one so far).   
   >   
   > --   
   > Regards,   
   > BB   
   >   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)