Sample Efficiency in Deep Reinforcement Learning based Recommender Systems with Imitation Learning
by Mohammad Mehdi Afsar, Trafford Crump, and Behrouz Far
Published onMay 27, 2022
Sample Efficiency in Deep Reinforcement Learning based Recommender Systems with Imitation Learning
·
Abstract
Recently, there has been an unprecedented interest in using reinforcement learning (RL) for recommender systems (RSs), due to its unique ability in taking into account the dynamic and long-term user engagement. However, sample ine ciency is a major challenge in applying RL to problems with very dynamic environments and huge actions spaces. In this paper, we present Imitation, Reinforcement learning based Recommender System (IR2S) to combine RL with imitation learning to alleviate this problem. More speci cally, by utilizing demonstrations (available user ratings), we show that IR2S can optimize its behavior faster and more e ciently. The proposed IR2S, built on top of Deep Q Network (DQN), shows superior performance compared to baselines in experiments.
Creative Commons Attribution 4.0 International License (CC-BY 4.0)