From: yaroslavvb@gmail.com   
      
   Hi, I'm looking for a way to increase the complexity of graphical model   
   one parameter at a time. The problem is that when I add full n'th order   
   interaction, I add a CPT with j^n-1 parameters, so that's a big jump in   
   complexity.   
      
   More formally, the problem is the following. Start with binary feature   
   vector v= and learn a probability distribution f:v->[0..1]   
   from data. To make this tractable, I start with fully disconnected   
   Graphical Model. Then I add some (hyper)edge, retrain, look at cross   
   validation error, and repeat. When cross-validation accuracy stops   
   increasing, terminate.   
      
   After I allowed all possible 4th order interactions, I may add a   
   hyperedge for 5th order interactions. But specifying that hyperedge as   
   a table will require me to specify 32 elements. That's a big jump in   
   complexity. What would be a cogent way to increase complexity one   
   parameter at a time? What assumptions does it make?   
      
   PS: A typical domain for this problem is Co-NLL named entity   
   recognition dataset.   
      
   Yaroslav Bulatov   
   bulatov@cs.oregonstate.edu   
   Dearborn 102   
   Oregon State University   
   Corvallis, OR   
      
   [ 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)