Skip to content

Comment on A gentle introduction to statistical relational learning

Comments

Can anyone explain to me how the "partition function" 'Z' is calculated? That part is a little glossed over in the post and the description in the paper is a little beyond my immediate understanding. :)

Partition functions are usually hard to compute as they require a sum or integral over the entire parameter space. In this case, it's "Z is a normalizing constant, that is: it's the sum of the values of all possible assignments."

So perhaps sum_x { exp( sum_i { w[i] * g[i](x) } ) }

tel is 100% correct.

And you're right, it is a little glossed over, I'll try to be more explicit. Thanks for the comment!

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.