I. User and item representations

• Enhancement of existing recommendation algorithms through deep learning methods

• Learning representations of items and/or users using multiple information sources

II. Dynamic behavior modeling

• Dynamic temporal user behavior modeling

• Session and intention modeling

III. Specialized recommendation methods using deep learning techniques

• Incorporating unstructured data sources such as text, audio, video or image into recommendation algorithms

• Context-aware recommender systems

• Handling the cold-start problem with deep learning

• Application specific deep learning based recommenders (e.g. music recommenders)

IV. Architecture

• Novel deep neural network architectures for a particular recommendation task

• Scalability of deep learning methods for real-time applications

• Advances in deep learning technology for large scale recommendation

• Special layers or units designed for recommender systems

• Special activation functions or operators designed for recommender systems

V. Novel evaluation and explanation techniques

• Evaluation and comparison of deep learning implementations for a recommendation task

• Modeling the state of the user

• Sensitivity analysis of the network architecture

• Explanation of recommendations based on deep learning