A recurrent neural networks (RNN) is a special kind of neural network for modeling sequences, and it is quite successful in a number applications. One such application is sequence generation. In the article The Unreasonable Effectiveness of Recurrent Neural Networks, Andrej Karpathy writes about multiple examples where RNNs show very impressive results, including generation of Shakespeare, Wikipedia articles, XML, Latex, and even C code!
Since they have proven useful in a few applications already, the natural question to ask is whether we can apply RNNs to some other domains. What about recommender systems, for example? This is the question the authors of the recurrent neural networks Based Subreddit Recommender System report have asked themselves (see https://cole-maclean.github.io/blog/RNN-Based-Subreddit-Recommender-System/). The answer is yes, we...
