collaborative filtering

Restricted Boltzmann Machines (RBMs) are an unsupervised learning method (like principal components). An RBM is a probabilistic and undirected graphical model. They are becoming more popular in machine learning due to recent success in training them with contrastive divergence. They have been proven useful in collaborative filtering, being one of the most successful methods in the Netflix challenge (paper). Furthermore, they have been tantamount to training deep learning models, which appear to be the best current models for image and digit recognition.

Introduction Matrix factorization has been proven to be one of the best ways to do collaborative filtering. The most common example of collaborative filtering is to predict how much a viewer will like a movie. The power of matrix factorization was a key development of the Netflix Prize (see http://www2.research.att.com/~volinsky/papers/ieeecomputer.pdf). Using the movie rating example, the idea is that there are some underlying features of the movie and underlying attributes of the user that interact to determine if the user will like the movie.