Gaussian Process Neural Network Embeddings for Collaborative Filtering

Wei Zhang, Brian Barr, John Paisley

Abstract: We propose a neural network approach to collaborative filtering based on the random Fourier feature approximation to the Gaussian process. The result is a neural network for regression consisting of two non-linear layers in which only the weights of the middle layer need to be learned, along with the embeddings of each user and object. The model is mathematically equivalent to a Gaussian process with RBF kernel on the space of concatenated user/object embeddings, with the advantage of the representation we use being its scalability and learnability using standard backpropagation methods. We demonstrate performance on several collaborative filtering datasets, the MovieLens data set, the Amazon data set and the Netflix challenge set, and compare favorably with several related deep and non-deep collaborative filters.