From: mlh496@gmail.com   
      
   Hey everyone,   
      
   I'm a big fan of graphical models.  It is my understanding that there   
   are 2 broad classes of graphical models: dynamic and static.  I have   
   begun to wonder if there could possibly be a third class: continuous.   
   Here is how I think it differs from dynamic models.  In the literature,   
   it seems that dynamic graphical models typically assume a fixed time   
   increment dt.  Now suppose you decide to go completely real-time where   
   instruments may provide data at fixed intervals, but there are other   
   exogenous events and data that arrive sporadically.  When constructing   
   the learning agent, does one typically pick a fine-enough dt and then   
   lump the event in the closest time slice?  Or are there models where   
   you use the data as it arrives and execute your training algorithm on   
   demand? (i.e. continuous)   
      
   I know there are extreme complexities with this approach.  In a dynamic   
   model you generally have 2 models: a static model and a transition   
   model.  If the gap between time slices is variable, then you may not be   
   able to supply a meaningful prior to the parameters of the transition   
   model.   
      
   If anyone can convince me this is a dead-end, or point me to some   
   literature that picks up on this topic, I'd be most appreciative.   
      
   -Michael   
      
   [ comp.ai is moderated ... your article may take a while to appear. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)