The effect of priors on Learning with Restricted Boltzmann Machines