A Survey of Recommender Systems Based on Deep Learning

Ruihui Mu

doi:10.1109/access.2018.2880197

Abstract

In recent years, deep learning’s revolutionary advances in speech recognition, image analysis, and natural language processing have gained significant attention. Deep learning technology has become a hotspot research field in the artificial intelligence and has been applied into recommender system. In contrast to traditional recommendation models, deep learning is able to effectively capture the non-linear and non-trivial user-item relationships and enables the codification of more complex abstractions as data representations in the higher layers. In this paper, we provide a comprehensive review of the related research contents of deep learning-based recommender systems. First, we introduce the basic terminologies and the background concepts of recommender systems and deep learning technology. Second, we describe the main current research on deep learning-based recommender systems. Third, we provide the possible research directions of deep learning-based recommender systems in the future. Finally, concludes this paper.

Highlights

In recent years, with the rapid development of sensor technology, storage technology, computer technology, and network technology, the explosion of data has become increasingly intensified [1]
Deep learning-based recommender systems(DLRS) are of such leading solutions to these challenges, which are appropriate tools to quickly aid the process of information seeking
Deep learning-based recommender systems can learn the latent representations of users and items from massive data, and construct a recommendation model, generate an effective recommendation list for the user

Summary

INTRODUCTION

With the rapid development of sensor technology, storage technology, computer technology, and network technology, the explosion of data has become increasingly intensified [1]. Collaborative filtering-based recommendation methods [6] makes full use of the behavior information and preference information generated by the user in the past without using the user’s personal information and product description information, such as the user’s rating of the item to generate the recommended item. Deep learning is able to effectively capture the non-linear and non-trivial user-item relationships and enable the codification of more complex abstractions as data representations in the higher layers [12]–[24]. It catches the intricate relationships within the data itself, from abundant accessible data sources such as contextual, textual and visual information.

TERMINOLOGIES AND BACKGROUND CONCEPTS

DEEP LEARNING-BASED HYBRID RECOMMENDER SYSTEMS

POSSIBLE RESEARCH DIRECTIONS

CONCLUSION