Re:A gentle introduction to statistical relational learning
by braaap on Wednesday July 15, 2015 @06:36PM
(#9894463)
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. :)
Re:A gentle introduction to statistical relational learning
by tel on Wednesday July 15, 2015 @06:54PM
(#9894520)
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) } ) }
And you're right, it is a little glossed over, I'll try to be more explicit. Thanks for the comment!
∀x:Add(x,0)=0
Whut? Shouldn't it be
∀x:Add(x,0)=x ?