Online Learning to Rank in Stochastic Click Models

TL;DR

This paper introduces BatchRank, an online learning to rank algorithm applicable to a broad class of click models, including cascade and position-based models, with proven regret bounds and superior empirical performance.

Contribution

We propose BatchRank, the first algorithm for online learning to rank that works across multiple click models with theoretical guarantees.

Findings

BatchRank outperforms ranked bandits in experiments.

BatchRank is more robust than CascadeKL-UCB.

Theoretical regret bounds are established for BatchRank.

Abstract

Online learning to rank is a core problem in information retrieval and machine learning. Many provably efficient algorithms have been recently proposed for this problem in specific click models. The click model is a model of how the user interacts with a list of documents. Though these results are significant, their impact on practice is limited, because all proposed algorithms are designed for specific click models and lack convergence guarantees in other models. In this work, we propose BatchRank, the first online learning to rank algorithm for a broad class of click models. The class encompasses two most fundamental click models, the cascade and position-based models. We derive a gap-dependent upper bound on the $T$-step regret of BatchRank and evaluate it on a range of web search queries. We observe that BatchRank outperforms ranked bandits and is more robust than CascadeKL-UCB, an existing algorithm for the cascade model.