[countersixte]

For a (slightly technical) in-depth guide to training RBM's: http://www.cs.toronto.edu/~hinton/absps/guideTR.pdf

It discusses how to choose a learning algorithm, selecting hyperparameters, number of hidden units, etc.