hw2sol - Probablistic Graphical Models, Spring 2007...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Probablistic Graphical Models, Spring 2007 Homework 2 solutions December 6, 2007 1 Markov Networks Solution due to Yongjin Park Strong Union: X Y | Z = X Y | Z,W (1) Transitivity: ( X A | Z ) ( A Y | Z ) = ( X Y | Z ) (2) 1. See Figure 1. Figure 1: BN G of which I ( G ) does not satisfy these two properties. 2. The global Markov assumption defines X Y | Z if and only if sep H ( X ; Y | Z ), or there is no other path between X and Y that does not pass through Z . Due to the monotonicity of separation, sep H ( X ; Y | Z ) = se H ( X ; Y | Z ) , (3) where Z Z and { X,Y } / Z . Then it follows that any set of nodes Z s.t. Z Z , X Y | Z = X Y | Z . Since { Z } { Z,W } , the strong union property holds in I ( H ). As for the transitivity, we can prove by showing the contrapositive proposition is correct: A, X Y | Z = X A | Z A Y | Z (4) There can be three cases of A without violating X Y | Z , A belongs to the only path between X and Y where sep H ( X ; Y | Z ). If A is between X and Z then sep H ( X ; A | Z ) but sep H ( A ; Y | Z ); thus A Y | Z . The mirror case is also true. A is connected to either X or Y , but not simultaneously. If a path between X and A , but not between Y and A , sep H ( A ; X | Z ) but sep H ( A ; Y | ). The mirror case is also true. Thus, either X A | Z or A Y | Z . A is not connected with either X or Y at all. Trivially, X A Y A . 1 3. Show I l ( H ) = I p ( H ). Equivalently we can prove by showing that if X Y | N H ( X ), then X Y |X { X,Y } , for all disjoint sets X,Y . By the definition of I-map, X Y | N H ( X ) sep H ( X ; Y | N H ( X )) . (5) Again exploiting the monotonicity of d separation (Eq. 3), if N H ( X ) Z , Z d-separates X and Y , i.e. sep H ( X ; Y | Z ). Inductively, adding each node A [ X { X,Y } N H ( X )] to Z , i.e. Z Z { A } , where Z is initially N H ( X ), sep H ( X ; Y | N H ( X )) = = sep H ( X ; Y | Z ) = sep H ( X ; Y | Z ) = = sep H ( X ; Y |X { X,Y } ) . (6) In sum, X Y | N H ( X ) = X Y |X { X,Y } , and I l ( H ) = I p ( H )....
View Full Document

This note was uploaded on 05/25/2008 for the course MACHINE LE 10708 taught by Professor Carlosgustin during the Fall '07 term at Carnegie Mellon.

Page1 / 6

hw2sol - Probablistic Graphical Models, Spring 2007...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online