variational Bayes approximation of Boltzmann machine (or restricted Boltzmann machine)?

I apologize if this is the wrong subreddit to post to, but I was wondering if anyone is familiar with a variational Bayes algorithm to perform inference for either a general Boltzmann machine, or on the restricted Boltzmann machine? Any help is greatly appreciated!

p.s. Also, has anyone seen a penalized Boltzmann machine (e.g. where the weights between layers have some sort of Lq norm on them? I've done a literature search, and can't seem to find anything along these lines.)