From: newsgroup@silveraxe.de   
      
   Hello,   
   I am working with modeling Bayesian networks for a Master thesis and I   
   have a problem understanding and handling dependency in Noisy-Or. As far   
   as I understood I can only use Noisy-Or if the predecessors are independent.   
      
   The problem I have is that the predecessors are not independent in my   
   network. Now I am trying to remodel the network to reach independent   
   predecessors. I have the following idea:   
      
   I took the concept of Adding-Up from Finn Jensen's "Bayesian Networks   
   and Decision Graphs". In this concept the case is explained when the   
   effect of the occurence of two events is different from the sum of the   
   single events (e.g. P(B|A1)=0.2 and P(B|A2)=0.2, but P(B|A1,A2)=0.5). In   
   that case I have dependency and I am wondering if I can model the   
   network the following way to still be able to use the Noisy-Or concept   
   and calculate the probability of P(B|A1-2,A3,...). I tried to show the   
   example in the following figure:   
      
   network with dependency   
      
        B   
       ^^^   
      / | \   
     A1 A2 A3...   
      
   to network with independence:   
      
        B   
      ^   ^   
     /     \   
   A1-2   A3   
     ^^   
   /  \   
   A1  A2   
      
      
      
   There is also a second question I have:   
      
   In the following model I cannot calculate P(C|A,B) with Noisy-Or because   
   A and B are not independent:   
      
        C   
       ^ ^   
      /   \   
     A --> B   
      
   Is it possible to change the network the following way to be able to use   
   the Noisy-Or concept?   
      
              C   
             ^ ^   
            /   \   
           B'    A   
           ^   
           |   
           A   
      
      
   It would be great if anybody can help me with this topic. Maybe somebody   
   can also provide me with some literature to support this theory in my   
   Master thesis.   
      
   If there is anything unclear because of my explanation please let me   
   know and I try to explain it further.   
      
   Thank you very much,   
   Hilmar   
      
   [ 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)