Variational Autoencoders for Collaborative Filtering

16 Feb 2018Dawen LiangRahul G. KrishnanMatthew D. HoffmanTony Jebara

We extend variational autoencoders (VAEs) to collaborative filtering for implicit feedback. This non-linear probabilistic model enables us to go beyond the limited modeling capacity of linear factor models which still largely dominate collaborative filtering research.We introduce a generative model with multinomial likelihood and use Bayesian inference for parameter estimation... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Collaborative Filtering Million Song Dataset Mult-VAE PR [email protected] 0.266 # 2
Collaborative Filtering Million Song Dataset Mult-VAE PR [email protected] 0.364 # 1
Collaborative Filtering Million Song Dataset Mult-DAE [email protected] 0.266 # 2
Collaborative Filtering Million Song Dataset Mult-DAE [email protected] 0.363 # 2
Collaborative Filtering MovieLens 20M Mult-VAE PR [email protected] 0.395 # 2
Collaborative Filtering MovieLens 20M Mult-VAE PR [email protected] 0.537 # 2
Collaborative Filtering MovieLens 20M Mult-DAE [email protected] 0.387 # 3
Collaborative Filtering MovieLens 20M Mult-DAE [email protected] 0.524 # 3
Collaborative Filtering Netflix Mult-VAE PR [email protected] 0.351 # 2
Collaborative Filtering Netflix Mult-VAE PR [email protected] 0.444 # 2
Collaborative Filtering Netflix Mult-DAE [email protected] 0.344 # 3
Collaborative Filtering Netflix Mult-DAE [email protected] 0.438 # 3