From: markmaloof@mail.com   
      
   Taking a weighted majority vote is only part of it.  Littlestone's   
   Weighted Majority algorithm predicts based on a weighted majority of the   
   expert predictions, but it also decreases the weights of experts that predict   
   incorrectly.   
      
   We have to be a little careful about what it means to be "an expert in one's   
   field."  Most learning algorithms that you'd use for experts assume that   
   future observations will be drawn from the same distribution as those   
   examples on which they were trained.   
      
   However, you might want to check out Avrim Blum's follow up to Littlestone's   
   work.  He used pairs and triplets of features as experts, so depending on   
   your definition of an expert's field and how far you want to take that   
   analogy, Blum's experts were experts and specialists in different regions   
   of the representation space.  For example, one useful expert might be good   
   at making predictions using color and shape, while another might be good   
   using size and weight.   
      
   We've looked at how to dynamically add and remove experts for problems   
   involving patterns that change over time (aka concept drift).   
      
   Hope this helps.   
      
   Mark   
      
   [ comp.ai is moderated.  To submit, just post and be patient, or if ]   
   [ that fails mail your article to , and ]   
   [ ask your news administrator to fix the problems with your system. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)