XPost: comp.ai.neural-nets   
   From: kimon@ath.forthnet.gr   
      
   > I don't agree that SMO is especially fast.  SMO is an excellent   
   > algorithm because it's easy to code up if you don't have a general   
   > purpose QP solver.  Having done a fair bit of experimentation and   
   > comparison, I don't feel that it's in general faster than using a   
   > state-of-the-art QP solver to solve moderate size (say a couple   
   > hundred) subproblems at each iteration.  I'm not sure what would   
   > happen if you tried to solve for 3 or 4 examples at once.   
   >   
      
   That's surprising for me because that is exactly the size of   
   sub-problem I thought SMO is better. Which QP solver have you used?   
   Surely you don't mean a general purpose one.   
      
   Kimon   
      
   [ 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)