Assignment 2 - Problem 1 - Markov Blanket

Moderator: Computer Vision 2

lustiz
Mausschubser
Mausschubser
Beiträge: 70
Registriert: 29. Apr 2009 10:28

Assignment 2 - Problem 1 - Markov Blanket

Beitrag von lustiz » 4. Dez 2012 09:13

Hey guys, does the Markov blanket for a variable include the variable itself? Wiki says it doesn't. However, I found a definition in the ML2 slides of Prof. Roth that includes the variable which we want to compute the blanket for.

robert.n
Nerd
Nerd
Beiträge: 673
Registriert: 29. Sep 2008 19:17

Re: Assignment 2 - Problem 1 - Markov Blanket

Beitrag von robert.n » 4. Dez 2012 13:10

lustiz hat geschrieben:Hey guys, does the Markov blanket for a variable include the variable itself? Wiki says it doesn't. However, I found a definition in the ML2 slides of Prof. Roth that includes the variable which we want to compute the blanket for.
No, it doesn't. If it could include the variable itself, then why include other variables at all? If you already know the value of a variable, then obviously the value of this variable is already independent of all other variables. Example: x1 is independent of all other variables, given x1.

This is also why it says "and all the other parents of its children".

Conclusion:
Many illustrations for the Markov blanket are wrong, including the illustration on wikipedia and in the Computer Vision book by Prince. However, there are also many correct illustrations. See Google Image search.

This illustration is correct: http://lecture.ecc.u-tokyo.ac.jp/~yamag ... lanket.png
This is not: https://origin-ars.els-cdn.com/content/ ... 5X-gr2.jpg

franzel
Erstie
Erstie
Beiträge: 21
Registriert: 6. Apr 2009 20:22

Re: Assignment 2 - Problem 1 - Markov Blanket

Beitrag von franzel » 5. Dez 2012 11:15

Robert is right regarding the definition.

The formal definitions from both books ("Probabilistic Graphical Models: Principles and Techniques", "Computer vision:models, learning and inference") suggest that the markov blanket of variable X does not include X itself.

However, it is convention (c.f. "Computer vision:models, learning and inference") to include the the node into the the shaded area when depicting the markov blanket.

Antworten

Zurück zu „Computer Vision 2“