I think I understand how to calculate BN and why it is so, but complex net still confuses me. Currently how I understand it is that, if there is any 'result' variable in the probability, it can be calculated by converting it to P(result|cause)P(cause). So I assumed in this bayesian network (A,B → C → D,E), if you want to get P(B,E), every parent and grand parent of B and E should be considered. As B doesn't have parent, only the (grand) parent of E (A,C) will hop in, and finally it can be calculated as :

P(B,E) = P(B,E,A,C)+P(B,E,~A,C)+P(B,E,A,~C)+P(B,E,~A,~C)

The joint probability would be :

P(B,E,A,C) = P(B)P(A)P(C|A,B)P(E|C)

The rest of them are calculated the same way. I'd like to check if this calculation and my understanding is right. Thank you!


For the DAG: $$\begin{array}{ccc}A&&&&B\\&\searrow&&\swarrow \\&&C\\&\swarrow&&\searrow\\D&&&&E\end{array}$$

Then, yes,$$P_{B,E}=\sum_{A,C}P_AP_BP_{C\mid A, B}P_{E\mid C}$$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.