[Paper Review] Deep Learning based Recommender System: A Survey and New Perspectives

Shuai Zhang, Lina Yao+2 more|arXiv (Cornell University)|Jul 24, 2017

Recommender Systems and Techniques204 references656 citations

TL;DR

A comprehensive survey of deep learning–based recommender systems, proposing a taxonomy, state-of-the-art overview, challenges, and future directions.

ABSTRACT

With the ever-growing volume of online information, recommender systems have been an effective strategy to overcome such information overload. The utility of recommender systems cannot be overstated, given its widespread adoption in many web applications, along with its potential impact to ameliorate many problems related to over-choice. In recent years, deep learning has garnered considerable interest in many research fields such as computer vision and natural language processing, owing not only to stellar performance but also the attractive property of learning feature representations from scratch. The influence of deep learning is also pervasive, recently demonstrating its effectiveness when applied to information retrieval and recommender systems research. Evidently, the field of deep learning in recommender system is flourishing. This article aims to provide a comprehensive review of recent research efforts on deep learning based recommender systems. More concretely, we provide and devise a taxonomy of deep learning based recommendation models, along with providing a comprehensive summary of the state-of-the-art. Finally, we expand on current trends and provide new perspectives pertaining to this new exciting development of the field.

Motivation & Objective

• Provide a systematic review of deep learning techniques used in recommender systems.
• Propose a taxonomy to organize existing deep learning–based recommendation models.
• Summarize state-of-the-art models and applications across data modalities and tasks.
• Discuss challenges, open issues, and future research directions in this area.

Proposed method

• Classify deep learning based recommender models by the type of neural building blocks (e.g., MLP, AE, CNN, RNN, RBM, NADE, Attentional Models, Adversarial Networks, DRL).
• Describe a second dimension for deep hybrid models that combine multiple neural components.
• Review a broad set of publications (over 100 studies) from major venues and databases to map current trends.
• Highlight example architectures such as Neural Collaborative Filtering and Deep Factorization Machines and how they extend traditional MF.

Experimental results

Research questions

• RQ1 What is the current landscape of deep learning techniques used in recommender systems?
• RQ2 How can we systematically categorize deep learning–based recommender models and what are the strengths/limitations of these categories?
• RQ3 What are the key challenges and open issues in deploying deep learning–based recommenders, and what future directions are promising?

Key findings

• Deep learning enables nonlinear interaction modeling, rich representation learning, sequence modeling, and multi-modal data integration for recommendations.
• A flexible end-to-end differentiable framework allows composing multiple neural building blocks into hybrid models applicable to various data sources (text, images, audio, video).
• RNNs and CNNs provide effective mechanisms for sequential and contextual recommendation tasks like session-based and next-item prediction.
• Attention mechanisms and neural building blocks contribute to improved interpretability and explainability in recommendations.
• The survey identifies open issues such as interpretability, data requirements, and hyperparameter tuning, and outlines future directions for the field.