Ranking, Reputation and Recommender Systems
Part III essay topic (2011-12)
To quote [1], "Ranking, reputation, recommendation, and trust systems have become essential ingredients of web-based multi-agent systems. These systems aggregate agents' reviews of one another, as well as about external events, into valuable information. Notable commercial examples include Google's page ranking system, and Amazon and E-bay's recommendation and reputation systems."
Your task is to read around this general field and write an interesting essay about mathematics for ranking, reputation and recommender systems. You can discuss both modelling and algorithms. You will need to be selective and focus on a few topics that you judge to be seminal (PageRank is an example), or where you find that current research is active, or where there are unsolved questions. The following references should get you started, but you can find other sources via your own explorations. I suggest you begin by reading [1] and [5].
Relevant Courses
Useful: Stochastic Networks, Mathematics of Operational Research
References
[1] Andersen, A. et al., Trust-based recommendation systems: an axiomatic approach, Microsoft Research, 2008.
[2] Gyongyi, Z, et al., Combating web spam with TrustRank. Proceeding VLDB '04 Proceedings of the Thirtieth international conference on Very large data bases - Volume 30, 2004.
[3] Keinberg, J. M., Authoritative sources in a hyperlinked environment, JACM 46, 1999.
[4] Page, L. PageRank: Bringing order to the Web. Stanford Digital Library Project, 1997.
[5] Renick, P. and Varian, H. R., Recommender systems,
[6] Tahajod, M. Iranmehr, A., Khozooyi, N., Trust management for semantic web, Computer and Electrical Engineering, 2009. ICCEE '09.
[7] Vojnovic, M., Cruise, J., Gunawardena, D., Marbach, P., Ranking and suggesting popular items, IEEE Trans. Knowledge and Data Engineering, 12, 1133--1146, 2009.
Additional references
The following reading has been suggested by Thore Graepel (Microsoft Research).About the TrueSkill Ranking System serving millions of gamers on Xbox Live
[8] Ralf Herbrich, Tom Minka, and Thore Graepel, TrueSkill(TM): A Bayesian Skill Rating System, in Advances in Neural Information Processing Systems 20, MIT Press, January 2007.
About the Matchbox Recommender System currently tested on users in Xbox Live
[9] David Stern, Ralf Herbrich, and Thore Graepel, Matchbox: Large Scale Bayesian Recommendations, in Proceedings of the 18th International World Wide Web Conference, 2009.
Here are a few other ideas for your reading, contributed by Felix Fischer.
He says, "Sep Kamvar has a few papers on personalized ranking: kamvar.org/personalized_search. The one on "The second eigenvalue of the Google matrix" is rather famous. Kamvar founded a startup based on the idea of personalized ranking that got bought up by Google, and he's a pretty interesting guy in general (other ventures include a fashion label and a piece acquired by the Museum of Modern Art that searches blogs for people's feelings and visualizes them)."
A CS theory paper on (approximation) algorithms for ranking and clustering: Aggregating inconsistent information: ranking and clustering
A paper in economics on web and citation rankings: The measurement of intellectual influence
Papers about axioms we would want a ranking method to satisfy, one by computer scientists and one by economists: Axiomatic foundations for ranking systems and Scoring of web pages and tournaments - axiomatizations
University of Cambridge > Mathematics > Statistical Laboratory > Richard Weber > Teaching page > Essays
Richard Weber ( rrw1@cam.ac.uk )
Last modified: 23 November 2011