July 2018 | SHUAI ZHANG, University of New South Wales; LINA YAO, University of New South Wales; AIXIN SUN, Nanyang Technological University; YI TAY, Nanyang Technological University
This article provides a comprehensive review of recent research efforts on deep learning-based recommender systems. It aims to offer a taxonomy of deep learning-based recommendation models and summarize the state-of-the-art. The authors discuss the challenges and open issues in the field, and identify new trends and future directions. The article highlights the advantages of deep learning in capturing nonlinear and complex user-item relationships, learning feature representations from scratch, and handling multi-modal data. It also addresses common limitations such as interpretability, data requirements, and hyperparameter tuning. The survey covers a wide range of deep learning techniques, including multilayer perceptrons (MLPs), autoencoders, convolutional neural networks (CNNs), recurrent neural networks (RNNs), and deep reinforcement learning (DRL). The authors provide detailed overviews of various deep learning-based recommendation models, such as neural collaborative filtering, deep factorization machines, wide & deep learning, and collaborative deep learning. The article concludes by discussing the potential of deep learning in enhancing user experience and promoting sales/services in various online applications.This article provides a comprehensive review of recent research efforts on deep learning-based recommender systems. It aims to offer a taxonomy of deep learning-based recommendation models and summarize the state-of-the-art. The authors discuss the challenges and open issues in the field, and identify new trends and future directions. The article highlights the advantages of deep learning in capturing nonlinear and complex user-item relationships, learning feature representations from scratch, and handling multi-modal data. It also addresses common limitations such as interpretability, data requirements, and hyperparameter tuning. The survey covers a wide range of deep learning techniques, including multilayer perceptrons (MLPs), autoencoders, convolutional neural networks (CNNs), recurrent neural networks (RNNs), and deep reinforcement learning (DRL). The authors provide detailed overviews of various deep learning-based recommendation models, such as neural collaborative filtering, deep factorization machines, wide & deep learning, and collaborative deep learning. The article concludes by discussing the potential of deep learning in enhancing user experience and promoting sales/services in various online applications.