English

Reinforcement Learning for Slate-based Recommender Systems: A Tractable Decomposition and Practical Methodology

Machine Learning 2019-06-03 v2 Artificial Intelligence Information Retrieval Machine Learning

Abstract

Most practical recommender systems focus on estimating immediate user engagement without considering the long-term effects of recommendations on user behavior. Reinforcement learning (RL) methods offer the potential to optimize recommendations for long-term user engagement. However, since users are often presented with slates of multiple items - which may have interacting effects on user choice - methods are required to deal with the combinatorics of the RL action space. In this work, we address the challenge of making slate-based recommendations to optimize long-term value using RL. Our contributions are three-fold. (i) We develop SLATEQ, a decomposition of value-based temporal-difference and Q-learning that renders RL tractable with slates. Under mild assumptions on user choice behavior, we show that the long-term value (LTV) of a slate can be decomposed into a tractable function of its component item-wise LTVs. (ii) We outline a methodology that leverages existing myopic learning-based recommenders to quickly develop a recommender that handles LTV. (iii) We demonstrate our methods in simulation, and validate the scalability of decomposed TD-learning using SLATEQ in live experiments on YouTube.

Keywords

Cite

@article{arxiv.1905.12767,
  title  = {Reinforcement Learning for Slate-based Recommender Systems: A Tractable Decomposition and Practical Methodology},
  author = {Eugene Ie and Vihan Jain and Jing Wang and Sanmit Narvekar and Ritesh Agarwal and Rui Wu and Heng-Tze Cheng and Morgane Lustman and Vince Gatto and Paul Covington and Jim McFadden and Tushar Chandra and Craig Boutilier},
  journal= {arXiv preprint arXiv:1905.12767},
  year   = {2019}
}

Comments

Short version to appear IJCAI-2019

R2 v1 2026-06-23T09:32:27.077Z