(ML 13.11) D-separation (part 2)

[@arifabd]

The HEAD TO HEAD relationship explained at 03:40 for the path from Node A to Node B is NOT BLOCKED with respect to node C. Because although node C interfaces a HEAD TO HEAD relation for the said path, node C is in set C, which are the same here.

[@LinchaoBao]

I think the "d" in d-separation stands for "dependence" but not for "directed".

[@christian.adriano]

Correct, follows reference - The "d" in d-separation and d-connection stands for dependence. www.andrew.cmu.edu/user/scheines/tutor/d-sep.html

[@mIsterCherif]

He means the vertices must not be in C so that the path from A to B is blocked by C. (at 3:40)

[@cybertron299]

Whats the D-separation theorem for marginal independence ? It was quite obvious in the 3 variable case. Great videos!!

[@edopizza]

At 3:40 it doesn't make sense. There is a contradiction with what is stated in the definition (b). The vertex IS in C.

[@samuelbarham8483]

He hadn't colored it in -- meaning we're not conditioning on it. He is at that moment, I think, intuitively defining the SET C as being the set of vertices we're conditioning on; here, the VERTEX c is not in the SET C, because we're not conditioning on the vertex c. He noted that the choice of names was unfortunate in that case.

[@marcosrodriguez2496]

Imagine you have the following graph {X1->X2; X3->X2; X4} (X4 has no outgoing or incoming edges). According to this definition the following statement is correct: "X1 and X3 are d-separated by X4." This seems silly as X4 doesn't do anything.

[@narical]

superb !!

[@cybertron299]

Sorry...just realized how to derive marginals from DAGs.