Content-based multimedia information retrieval: State of the art and challenges
ACM Transactions on Multimedia Computing, Communications, and Applications (TOMCCAP)
Robust Face Recognition via Sparse Representation
IEEE Transactions on Pattern Analysis and Machine Intelligence
Transfer learning for collaborative filtering via a rating-matrix generative model
ICML '09 Proceedings of the 26th Annual International Conference on Machine Learning
Can movies and books collaborate?: cross-domain collaborative filtering for sparsity reduction
IJCAI'09 Proceedings of the 21st international jont conference on Artifical intelligence
Content-based recommendation systems
The adaptive web
Proceedings of the fourth ACM conference on Recommender systems
Multi-label boosting for image annotation by structural grouping sparsity
Proceedings of the international conference on Multimedia
Recommender Systems Handbook
Improving Recommender Systems by Incorporating Social Contextual Information
ACM Transactions on Information Systems (TOIS)
Improving Aggregate Recommendation Diversity Using Ranking-Based Techniques
IEEE Transactions on Knowledge and Data Engineering
Hi-index | 0.00 |
Traditional tensor factorization based context-aware collaborative filtering considers the context as homogeneous ones, which uses vectorization to implement the factorization as the single context version while ignoring many structural interactions between the heterogeneous contexts. However, cross media data in digital libraries have common and distinctive context, which can be used to discover the latent structural grouping semantics to improve the diversity of recommendation. In this paper, we propose a structural context-aware feature selection framework for cross media recommendation. Firstly, the TUCKER based tensor factorization is conducted on the N-dimensional user-item-content tensor. Then the hidden structural representation are defined as the solution of the structural sparse coding with the loss function by regularizing the terms according to some principle context components, which are optimally selected by the structural grouping sparsity (MtBGS) method. Finally, the top n items with the highest n prediction probabilities are recommended for specific user. Experiments conducted on a cross media dataset based on Douban.com show the effectiveness of diversity for cross media recommendation.