Hierarchical Reinforcement Learning for Integrated Recommendation

• DOI: 10.1609/aaai.v35i5.16580
• Publication Date: 2022-09-08T18:34:41Z
• Authors: Ruobing Xie; Shaoliang Zhang; Rui Wang; Feng Xia; Leyu Lin
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

Integrated recommendation aims to jointly recommend heterogeneous items in the main feed from different sources via multiple channels, which needs capture user preferences on both item and channel levels. It has been widely used practical systems by billions of users, while few works concentrate integrated systematically. In this work, we propose a novel Hierarchical reinforcement learning framework for (HRL-Rec), divides into two tasks channels sequentially. The low-level agent is selector, generates personalized list. high-level an recommender, recommends specific under constraints. We design various rewards accuracy diversity, four losses fast stable model convergence. also conduct online exploration sufficient training. experiments, extensive offline experiments billion-level real-world dataset show effectiveness HRL-Rec. HRL-Rec deployed WeChat Top Stories, affecting millions users. source codes are released https://github.com/modriczhang/HRL-Rec.