1
$\begingroup$

Given the probability: $$p(a,b,c)=p(a)p(c|a)p(b|c)$$

How to prove $$ p(a,b)=p(a)\sum_{c}p(c|a)p(b|c)=p(a)p(b|a) $$

Does anyone have idea how to prove the above sequence? I am confused how is the sum rule applied.

$\endgroup$
  • 1
    $\begingroup$ This is not true in general. If $c$ and $b$ are independent, you'll have $p(b) = p(b|a)$, which implies than $a$ and $b$ are independent, too. The formula is correct if and only if $b$ depends on $a$ through $c$, that is $p(b|a,c) = p(b|c)$. $\endgroup$ – Artem Sobolev Apr 25 '15 at 11:07
  • $\begingroup$ @Barmaley.exe Sorry I haven't given all conditions. I editted once again. Would you please give me some inspiration? Thanks a lot^_^ $\endgroup$ – spacegoing Apr 25 '15 at 11:18
  • 1
    $\begingroup$ Ok, then just note that $p(a,b,c) = p(b,c|a) p(a) $ (by definition) which leads to $ p(b,c|a) = p(c|a) p(b|c)$ (from the first statement). The right-hand side of the last equation is what you have in the sum, so the sum is equal to $\sum_c p(b,c|a)$. This is marginalization over $c$. $\endgroup$ – Artem Sobolev Apr 25 '15 at 11:24
  • $\begingroup$ Thats really helpful, thanks a lot! Would you please tell me how did you come up with that idea! I was looking at this like half hour notching my head :P $\endgroup$ – spacegoing Apr 25 '15 at 11:30
1
$\begingroup$

With multiple applications of the conditional probabilty rule you get:

$$p(a,b) = \sum_{c} p(a,b,c) = \sum_{c} p(b,c|a) p(a) = p(a) \sum_{c} p(b,c|a) = p(a) \sum_{c} p(b|a,c) p(c|a).$$

This is as far as you can actually go without more assumptions. In the special case where $b \text{ } \bot \text{ } a | c$ (i.e., $b$ is conditionally independent of $a$ given $c$) you have $p(b|a,c) = p(b|c)$ which gives you the desired result.

$\endgroup$

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.