restricted boltzmann machines (rbms)

RBMs are generative neural network models that can learn a probability distribution over their set of inputs. They're used in dimensionality reduction, classification, regression, collaborative filtering, and even topic modelling.