I've been studying a paper on Recommender systems using collaborative deep learning where in a hierarchical bayesian model called collaborative deep learning is developed.

Stacked denoising autoencoders are used for the feature learning of the input which is the data from citeulike. And CF is used for learning the content information and collaborative filtering for the ratings(feedback) matrix.

Here is a link to the paper :http://www.wanghao.in/paper/KDD15_CDL.pdf