Having read "Restricted Boltzmann Machines for Collaborative Filtering" (Salakhutdinov et. al. 2007), I'm wondering if there has been any follow-up work on applying graphical and/or deep architectures for recommendation engines.

The paper suggests some interesting possible extensions, such as using stacked RBMs instead of single-layer ones. However, I didn't find any follow-up papers by Salakhutdinov's publication page.


There are some publications using auto encoders instead of RBMs. This article is really interesting: https://www.nicta.com.au/pub-download/full/8604/

There is also works on content-based recommendation using deep learning techniques such as this from spotify: http://benanne.github.io/2014/08/05/spotify-cnns.html

Another very recent (2016) publication is "Deep Neural Networks for YouTube Recommendations" https://research.google.com/pubs/pub45530.html - It feeds a deep network with user activies and video embeddings and dense features as well

Those works could lead you to great references also.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.