Multi-dueling bandits and their application to online ranker evaluation

TY - GEN

T1 - Multi-dueling bandits and their application to online ranker evaluation

AU - Brost, Brian

AU - Seldin, Yevgeny

AU - Cox, Ingemar Johansson

AU - Lioma, Christina

N1 - Conference code: 25

PY - 2016/10/24

Y1 - 2016/10/24

N2 - Online ranker evaluation focuses on the challenge of efficiently determining, from implicit user feedback, which ranker out of a finite set of rankers is the best. It can be modeled by dueling bandits, a mathematical model for online learning under limited feedback from pairwise comparisons. Comparisons of pairs of rankers is performed by interleaving their result sets and examining which documents users click on. The dueling bandits model addresses the key issue of which pair of rankers to compare at each iteration. Methods for simultaneously comparing more than two rankers have recently been developed. However, the question of which rankers to compare at each iteration was left open. We address this question by proposing a generalization of the dueling bandits model that uses simultaneous comparisons of an unrestricted number of rankers. We evaluate our algorithm on standard large-scale online ranker evaluation datasets. Our experimentals show that the algorithm yields orders of magnitude gains in performance compared to state-of-the-art dueling bandit algorithms.

AB - Online ranker evaluation focuses on the challenge of efficiently determining, from implicit user feedback, which ranker out of a finite set of rankers is the best. It can be modeled by dueling bandits, a mathematical model for online learning under limited feedback from pairwise comparisons. Comparisons of pairs of rankers is performed by interleaving their result sets and examining which documents users click on. The dueling bandits model addresses the key issue of which pair of rankers to compare at each iteration. Methods for simultaneously comparing more than two rankers have recently been developed. However, the question of which rankers to compare at each iteration was left open. We address this question by proposing a generalization of the dueling bandits model that uses simultaneous comparisons of an unrestricted number of rankers. We evaluate our algorithm on standard large-scale online ranker evaluation datasets. Our experimentals show that the algorithm yields orders of magnitude gains in performance compared to state-of-the-art dueling bandit algorithms.

KW - cs.IR

KW - cs.LG

KW - stat.ML

U2 - 10.1145/2983323.2983659

DO - 10.1145/2983323.2983659

M3 - Konferencebidrag i proceedings

T3 - ACM International Conference on Information and Knowledge Management

SP - 2161

EP - 2166

BT - Proceedings of the 25th ACM International Conference on Information and Knowledge Management

PB - Association for Computing Machinery

T2 - 25th ACM International Conference on Information and Knowledge Management

Y2 - 24 October 2016 through 28 October 2016

ER -