From: ted.dunning@gmail.com   
      
   Mackay's book has a good section on how EM methods can go seriously   
   astray.  The basic problem (taking a Gaussian mixture example) is that   
   if any cluster decides to focus on only a single isolated point, it can   
   fit that point infinitely well.  Since all points are isolated on the   
   correct scale, this is an inherent problem with maximum likelihood   
   clustering.   
      
   One repair is to use a prior distribution on the parameters of the   
   clusters and then maximize the posterior likelihood (MAP instead of   
   ML).  Mackay discusses this and other options.   
      
   See http://www.inference.phy.cam.ac.uk/itprnn/book.html   
      
   [ 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)