Clustering and Inference From Pairwise Comparisons

Given a set of pairwise comparisons, the classical ranking problem computes a single ranking that best represents the preferences of all users. In this paper, we study the problem of inferring individual preferences, arising in the context of making personalized recommendations. In particular, we assume users form clusters; users of the same cluster provide similar pairwise comparisons for the items according to the Bradley-Terry model. We propose an efficient algorithm to estimate the preference for each user: first, compute the net-win vector for each user using the comparisons; second, cluster the users based on the net-win vectors; third, estimate a single preference for each cluster separately. We show that the net-win vectors are much less noisy than the high dimensional vectors of pairwise comparisons, therefore our algorithm can cluster the users reliably. Moreover, we show that, when a cluster is only approximately correct, the maximum likelihood estimation for the Bradley-Terry model is still close to the true preference.

[1]  Mark Braverman,et al.  Noisy sorting without resampling , 2007, SODA '08.

[2]  David Gross,et al.  Recovering Low-Rank Matrices From Few Coefficients in Any Basis , 2009, IEEE Transactions on Information Theory.

[3]  David C. Parkes,et al.  Generalized Method-of-Moments for Rank Aggregation , 2013, NIPS.

[4]  Paul N. Bennett,et al.  Pairwise ranking aggregation in a crowdsourced setting , 2013, WSDM.

[5]  R. Srikant,et al.  Jointly clustering rows and columns of binary matrices: algorithms and trade-offs , 2013, SIGMETRICS '14.

[6]  R. A. Bradley,et al.  RANK ANALYSIS OF INCOMPLETE BLOCK DESIGNS , 1952 .

[7]  Moni Naor,et al.  Rank aggregation methods for the Web , 2001, WWW '01.

[8]  Devavrat Shah,et al.  What's your choice? Learning the mixed multi-nomial logit model , 2014, SIGMETRICS 2014.

[9]  Joachim M. Buhmann,et al.  Cluster analysis of heterogeneous rank data , 2007, ICML '07.

[10]  Emmanuel J. Candès,et al.  The Power of Convex Relaxation: Near-Optimal Matrix Completion , 2009, IEEE Transactions on Information Theory.

[11]  D. Hunter.  MM algorithms for generalized Bradley-Terry models , 2003 .

[12]  Laurent Massoulié,et al.  Distributed user profiling via spectral methods , 2010, SIGMETRICS '10.

[13]  Yu Lu,et al.  Individualized rank aggregation using nuclear norm regularization , 2014, 2015 53rd Annual Allerton Conference on Communication, Control, and Computing (Allerton).

[14]  R. Duncan Luce,et al.  Individual Choice Behavior , 1959 .

[15]  John Guiver,et al.  Bayesian inference for Plackett-Luce ranking models , 2009, ICML '09.

[16]  Devavrat Shah,et al.  Learning Mixed Multinomial Logit Model from Ordinal Data , 2014, NIPS.

[17]  Arun Rajkumar,et al.  A Statistical Convergence Perspective of Algorithms for Rank Aggregation from Pairwise Data , 2014, ICML.

[18]  Joel A. Tropp,et al.  User-Friendly Tail Bounds for Sums of Random Matrices , 2010, Found. Comput. Math..

[19]  R. A. Bradley,et al.  RANK ANALYSIS OF INCOMPLETE BLOCK DESIGNS THE METHOD OF PAIRED COMPARISONS , 1952 .

[20]  Bruce E. Hajek,et al.  Minimax-optimal Inference from Partial Rankings , 2014, NIPS.

[21]  Onkar Dabeer,et al.  Analysis of a Collaborative Filter Based on Popularity Amongst Neighbors , 2012, IEEE Transactions on Information Theory.

[22]  Devavrat Shah,et al.  Rank Centrality: Ranking from Pairwise Comparisons , 2012, Oper. Res..

[23]  Yehuda Koren,et al.  Matrix Factorization Techniques for Recommender Systems , 2009, Computer.

[24]  Yuan Yao,et al.  Statistical ranking and combinatorial Hodge theory , 2008, Math. Program..

[25]  Anil N. Hirani,et al.  Least Squares Ranking on Graphs , 2010, 1011.1716.

[26]  MassouliéLaurent,et al.  Clustering and Inference From Pairwise Comparisons , 2015 .

[27]  David C. Parkes,et al.  Computing Parametric Ranking Models via Rank-Breaking , 2014, ICML.

[28]  Nebojsa Jojic,et al.  Efficient Ranking from Pairwise Comparisons , 2013, ICML.