Implementing Deep Neural Networks for YouTube Recommendations in PyTorch (in progress)

Deep Neural Networks for YouTube Recommendations, published in 2016 by Paul Covington, Jay Adams and Emre Sargin, has a great impact on the development of recommendation system. The system has two major components: candidate generation and ranking.

In this tutorial, I implement the candidate generation system. As YouTube does not release its data, I use MovieLens data set to model it.

Table of contents:

1. Import data sets. (Finished)
2. Data preprocessing. (Finished)
3. Model building. (Finished)
4. Model training. (Yet to finish)
5. Model evaluation. (Yet to finish)