I have a joint probability, which factors as follows:

$P(A,B,C,D) = P(A,B) \cdot P(C|A) \cdot P(D|B)$

So I know that $C$ and $D$ are independent given $P(A, B)$ right?

I want to infer $P(A,B|C,D)$.

I use Bayes' Rule:

$P(A,B,C,D) = P(A,B) \frac{P(A|C) \cdot P(C) \cdot P(B|D) \cdot P(D)}{P(A) \cdot P(B)}$

Then I condition on $P(C,D)$:

$P(A,B|C,D) = P(A,B) \frac{P(A|C) \cdot P(C) \cdot P(B|D) \cdot P(D)}{P(A) \cdot P(B) \cdot P(C,D)}$

My problem is, I do not know the terms $P(C)$ and $P(D)$ nor $P(C,D)$. What additional assumption would I have to make so that I can write:

$P(A,B|C,D) = P(A,B) \frac{P(A|C) \cdot P(B|D)}{P(A) \cdot P(B)}$


$P(A,B|C,D) \propto P(A,B) \frac{P(A|C) \cdot P(B|D) }{P(A) \cdot P(B)}$

Would it help if I assume, that both $P(A)$ and $P(B)$ are equally distributed? I guess, that then I at least can write the latter one, right?

Any help is very much appreciated! Thank you very much!


So I know that $C$ and $D$ are independent given $P(A,B)$ right?

They are independent given $A,B$, not $P(A,B)$.

I use Bayes' Rule: ...

Generally, you use Bayes' rule to write something in terms of things you do know. Why not start with the equation you wrote before that: $P(A,B,C,D) = P(A,B) \cdot P(C|A) \cdot P(D|B)$?


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.