Skip to yearly menu bar Skip to main content


Poster

UniRank: Unimodal Bandit Algorithms for Online Ranking

Camille-Sovanneary GAUTHIER · Romaric Gaudel · Elisa Fromont

Hall E #633

Keywords: [ T: Online Learning and Bandits ] [ MISC: Online Learning, Active Learning and Bandits ]


Abstract:

We tackle, in the multiple-play bandit setting, the online ranking problem of assigning L items to K predefined positions on a web page in order to maximize the number of user clicks. We propose a generic algorithm, UniRank, that tackles state-of-the-art click models. The regret bound of this algorithm is a direct consequence of the pseudo-unimodality property of the bandit setting with respect to a graph where nodes are ordered sets of indistinguishable items. The main contribution of UniRank is its O(L/∆ logT) regret for T consecutive assignments, where ∆ relates to the reward-gap between two items. This regret bound is based on the usually implicit condition that two items may not have the same attractiveness. Experiments against state-of-the-art learning algorithms specialized or not for different click models, show that our method has better regret performance than other generic algorithms on real life and synthetic datasets.

Chat is not available.