AMRank: An adversarial Markov ranking model combining short- and long-term returns

作者：

Highlights：

• AMRank, a novel document ranking model, is proposed, which combines MDP and GAN.

• AMRank employs long- and short-term returns to improve decision making.

• A sequence discriminator is presented to generate long-term returns.

• AMRank can realize one-step update and output return with less variance.

摘要

•AMRank, a novel document ranking model, is proposed, which combines MDP and GAN.•AMRank employs long- and short-term returns to improve decision making.•A sequence discriminator is presented to generate long-term returns.•AMRank can realize one-step update and output return with less variance.

论文关键词：Document ranking,Learning to rank,Reinforcement learning,Discriminator

论文评审过程：Received 27 November 2021, Revised 24 July 2022, Accepted 9 August 2022, Available online 22 August 2022, Version of Record 30 August 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118512