Reinforcement Learning for Slate-based Recommender Systems: A Tractable Decomposition and Practical Methodology
To read later
link