Collaborative Filtering Based on a Variational Gaussian Mixture Model

Collaborative filtering (CF) is a widely used method in recommendation systems. Linear models are still the mainstream of collaborative research methods, but non-linear probabilistic beyond limit linear model capacity. For example, variational autoencoders (VAEs) have been extensively CF, and achieved excellent results. Aiming at problem prior distribution for latent codes VAEs traditional CF too simple, which makes implicit variable representations users items poor. This paper proposes autoencoder that uses Gaussian mixture factors GVAE-CF. On this basis, an optimization function suitable GVAE-CF proposed. In our experimental evaluation, we show performance outperforms previously proposed VAE-based on several popular benchmark datasets terms recall normalized discounted cumulative gain (NDCG), thus proving effectiveness algorithm.