Under what conditions children of a node in Bayesian network are independent if the probability of the parent is unknown?

1.1k Views Asked by At

Given the following graph:

$$\lower{4ex}X\lower{2ex}\swarrow Y\lower{2ex}\searrow\lower{4ex}Z$$

define the probabilities such that Z and X are independent while the value of Y is unknown and prove that Z and X are indeed independent.

I know that if X and Z are independent then $P(X,Z)=P(X)\cdot P(Z)$. Also a node is conditionally independent of its non-descendants given its parents. But I'm not sure how to proceed with this.

1

There are 1 best solutions below

4
On BEST ANSWER

I know that if X and Z are independent then $P(X,Z)=P(X)⋅P(Z)$. Also a node is conditionally independent of its non-descendants given its parents. But I'm not sure how to proceed with this.

Yes, indeed, the outline is :

$$\begin{align}\mathsf P(X,Z) & =\sum_Y \mathsf P(X,Y,Z) \\[0ex]&~~\vdots \\[1ex]&=\sum_Y \mathsf P(X)\,\mathsf P(Y\mid X,Z)\,\mathsf P(Z) \\[1ex]&= \mathsf P(X)\,\mathsf P(Z)\sum_Y\mathsf P(Y\mid X,Z) \\[1ex]&= \mathsf P(X)\,\mathsf P(Z) \end{align}$$

You just need to justify that step.