« Boo! | Main | Who Took the Bomb? »
November 1, 2006
Awesome
First I posted some code as an excuse for having no post written, then I discovered a bug in the code, so now I have nothing at all. But check it out: factor graphs are awesome. They're more expressive than Bayes Nets and Markov Random Fields combined! Oooooh.
Posted by todd at November 1, 2006 11:20 PM
Comments
Post teh code and give $1 to each bug solved by your astute reading audience.
Posted by: jon at November 2, 2006 4:43 AM
Hm. Did I say 'bug'? I meant 'incomplete algorithm.'
Posted by: todd. at November 2, 2006 7:50 AM
Factor graphs aren't really more expressive than (undirected) Bayes Nets. There's a way to turn a Bayes Net into a factor graph (i.e. find a good ordering for elimination), and that's how the junction tree algorithm works.
Posted by: Leon at November 10, 2006 2:34 PM
My understanding -- which is undoubtedly not as full as yours -- was that the graphical model equivalent to "Bayes Nets if the links are undirected" was MRFs, and that there are functions expressible in FGs not expressible in BNs or MRFs. Now I have to go look up undirected bayes nets and the junction tree algorithm. Thanks.
Posted by: todd. at November 10, 2006 6:46 PM
Yes, you're right, as you explained to me in Pittsburgh. Factor graphs are definitely more expressive.
Posted by: Leon Barrett at November 28, 2006 1:36 PM
Post a comment
Thanks for signing in, . Now you can comment. (sign out)
(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)