Deep Federated Recommendation System
Keywords: recommender systems, neural collaborative filtering, federated learning
TL;DR: This paper studies the application of a federated learning approach to neural collaborative filtering for recommender systems.
Abstract: With the rapid development of deep learning algorithms in recent years, intelligent systems now play an intrinsic role in most of today’s industries. Consequently, there is high demand in deep-learning based applications which seek patterns in the user data to maximize the user experience and minimize the costs of a company. One such application is recommendation systems, which allow companies to market specific products directly to individuals that have the highest probability of being interested in the said product. In recent years we have seen a shift from primitive recommendation algorithms, such as collaborative filtering and matrix factorization to more complex deep learning approaches, which are capable of more effectively processing large volumes of data, while also exhibiting better performance. Such recommendation systems are employed by industry giants from virtually every field, such as digital marketing, streaming, video game industry and many more. However, with the increase in volumes of data, the concern for privacy also increases, and recently we have seen many cases where companies fail to protect the data of their consumers. In this report we propose a deep learning-based recommendation system that provides recommendations to a user based on their previous activity. To account for the increasing concern of data privacy, we will be employing a federated learning approach, using which we prevent users from directly sending their data to the server, avoiding any risk of it being intercepted by third parties.
Submission Number: 9
Loading