Figure 8: Phase-of-firing coding of natural visual stimuli in primary visual cortex. This figure was created in the same way as Figure 7. It shows that a less popular article might also have a high offset value j . In this case, it changes the actual magnitudes in θj , but does not bring in other topics.
5. CONCLUSIONS AND FUTURE WORK
We proposed an algorithm for recommending scientific articles to users based on both content and other users’ ratings. Our study showed that this approach works well relative to traditional matrix factorization methods and makes good predictions on completely unrated articles. Further, our algorithm provides interpretable user profiles. Such profiles could be useful in real-world recommender systems. For example, if a particular user recognizes her profile as representing different topics, she can choose to “hide” some topics when seeking recommendations about a subject.
Acknowledgements. The authors thank anonymous reviewers for their insightful comments. Chong Wang is supported by Google PhD fellowship. David M. Blei is supported by ONR 175-6343, NSF CAREER 0745520, AFOSR 09NL202, the Alfred P. Sloan foundation, and a grant from Google. 6.
REFERENCES
[1] D. Agarwal and B.-C. Chen. Regression-based latent factor models. In Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining, pages 19–28, New York, NY, USA, 2009. ACM.
[2] D. Agarwal and B.-C. Chen. flda: matrix factorization through latent Dirichlet allocation. In Proceedings of the third ACM international conference on Web search and data mining, WSDM ’10, pages 91–100, New York, NY, USA, 2010. ACM.