Overlay Management for Fully Distributed User-Based Collaborative Filtering
Offering personalized recommendation as a service in fully distributed applications such as file-sharing, distributed search, social networking, P2P television, etc, is an increasingly important problem. In such networked environments recommender algorithms should meet the same performance and reliability requirements as in centralized services. To achieve this is a challenge because a large amount of distributed data needs to be managed, and at the same time additional constraints need to be taken into account such as balancing resource usage over the network. In this paper we focus on a common component of many fully distributed recommender systems, namely the overlay network. We point out that the overlay topologies that are typically defined by node similarity have highly unbalanced degree distributions in a wide range of available benchmark datasets: a fact that has important—but so far largely overlooked—consequences on the load balancing of overlay protocols. We propose algorithms with a favorable convergence speed and prediction accuracy that also take load balancing into account. We perform extensive simulation experiments with the proposed algorithms, and compare them with known algorithms from related work on well-known benchmark datasets.
KeywordsLoad Balance Recommender System Benchmark Dataset Overlay Network Mean Absolute Error
Unable to display preview. Download preview PDF.
- 1.Garbacki, P., Iosup, A., Doumen, J., Roozenburg, J., Yuan, Y., Brinke, T.M., Musat, L., Zindel, F., van der Werf, F., Meulpolder, M., et al.: Tribler protocol specificationGoogle Scholar
- 4.Pitsilis, G., Marshall, L.: A trust-enabled P2P recommender system. In: Proc. 15th IEEE Intl. Workshops on Enabling Technologies: Infrastructure for Collaborative Enterprises (WETICE 2006), pp. 59–64 (2006)Google Scholar
- 7.Billsus, D., Pazzani, M.J.: Learning collaborative information filters. In: Proc. 15th Intl. Conf. on Machine Learning (ICML 1998), pp. 46–54. Morgan Kaufmann, San Francisco (1998)Google Scholar
- 9.Takács, G., Pilászy, I., Németh, B., Tikk, D.: Scalable collaborative filtering approaches for large recommender systems. Journal of Machine Learning Research 10, 623–656 (2009)Google Scholar
- 10.Lawrence, N.D., Urtasun, R.: Non-linear matrix factorization with gaussian processes. In: Proc. 26th Annual Intl. Conf. on Machine Learning (ICML 2009), pp. 601–608. ACM, New York (2009)Google Scholar
- 11.O‘Connor, M., Herlocker, J.: Clustering items for collaborative filtering. In: Workshop on Recommender Systems at 22nd ACM SIGIR (1999)Google Scholar
- 16.Bickson, D., Malkhi, D., Zhou, L.: Peer-to-Peer rating. In: Proc. 7th IEEE Intl. Conf. on Peer-to-Peer Computing, 2007 (P2P 2007), pp. 211–218. IEEE Computer Society, Los Alamitos (2007)Google Scholar
- 19.Pouwelse, J., Yang, J., Meulpolder, M., Epema, D., Sips, H.: Buddycast: an operational peer-to-peer epidemic protocol stack. In: Proc. 14th Annual Conf. of the Advanced School for Computing and Imaging, ASCI, pp. 200–205 (2008)Google Scholar
- 22.Akavipat, R., Wu, L.S., Menczer, F., Maguitman, A.: Emerging semantic communities in peer web search. In: Proc. Intl. workshop on Information retrieval in peer-to-peer networks (P2PIR 2006), pp. 1–8. ACM, New York (2006)Google Scholar
- 23.Ziegler, C.N., McNee, S.M., Konstan, J.A., Lausen, G.: Improving recommendation lists through topic diversification. In: Proc. 14th Intl. Conf. on WWW, pp. 22–32. ACM, New York (2005)Google Scholar
- 27.Jelasity, M., Montresor, A., Jesi, G.P., Voulgaris, S.: The Peersim simulator, http://peersim.sf.net