Deep Neural Networks for Recommender Systems

Abstract

The moment you log into any web service application on your devices, it tells you what you must watch, eat, buy, learn, or wear. Recommender systems have become the quintessential "Friend-Philosopher-Guide" guiding us through our digital lives. This chapter aims to look at recommender systems as one of the most popular and emerging applications of neural networks and how they wire and mesh up things and people to facilitate valuable personalized recommendations. These recommendations may be based on either user's behavior over the Internet or mapping user's behavior with other people of similar taste. The focus is on explaining the need of these systems, understanding the recommendation systems in practice, and identifying how they are currently mapped to users' fractionally changing preferences. A walkthrough in building a deep neural network along with the architecture of most popular neural network algorithms for recommendation systems will be given. The chapter will discuss the benefits of deep learning over machine learning for recommender systems, tuning and optimizing the hyperparameters of the deep neural network, and also throw light on the open issues in recommender systems for researchers.