With the ever-growing volume, complexity and dynamicity of onlineinformation, recommender system has been an effective key solution to overcomesuch information overload. In recent years, deep learning's revolutionaryadvances in speech recognition, image analysis and natural language processinghave gained significant attention. Meanwhile, recent studies also demonstrateits effectiveness in coping with information retrieval and recommendationtasks. Applying deep learning techniques into recommender system has beengaining momentum due to its state-of-the-art performances and high-qualityrecommendations. In contrast to traditional recommendation models, deeplearning provides a better understanding of user's demands, item'scharacteristics and historical interactions between them. This article aims to provide a comprehensive review of recent researchefforts on deep learning based recommender systems towards fosteringinnovations of recommender system research. A taxonomy of deep learning basedrecommendation models is presented and used to categorize the surveyedarticles. Open problems are identified based on the analytics of the reviewedworks and potential solutions discussed.
展开▼
