DRBM Implementation

Hey Everyone,

I am implementing a DRBM following this paper: http://machinelearning.org/archive/icml2008/papers/601.pdf

I am stuck on p(y | h), the difficulty I am having is with dy and dy^, what would be examples of those values? I know that d represents the biases of the class units, but what exactly would dy^ be? I speculate that dy could simply be the product of d and c , where c is the desired class and I am thinking dy^ could be the sum of the products of d and all possible c.

All that aside, I have also started looking at this now: https://dslpitt.org/uai/papers/11/p463-louradour.pdf

And in this paper, p(y=ec | h) has a different form. Which would you recommend? And why are they different?

Thank you!