报告题目：On the Opportunities and Challenges of Offline Reinforcement Learning for Recommender Systems
报告专家：姚丽娜 教授
报告时间：2024年6月12日（周三） 下午16:00
报告地点：腾讯会议（757-175-826）
主持人：刘琦 教授、许小龙 教授
专家简介：
Prof. Lina Yao, Data 61, CSIRO, Sydney, Australia.
Lina Yao is a Senior Principal Research Scientist and Science Lead (Research Manager) @ CSIRO's Data61, Conjoint Professor @ University of New South Wales, Honorary Professor @ Macquarie University, and Adjunct Professor @ University of Technology Sydney. She is the Senior Members of ACM and IEEE. Her research interest includes Few-Shot Learning, Zero-Shot Learning, Deep Reinforcement Learning, Self-supervised Learning, Deep Generative Modelling and their applications in a broad range of applications in Recommender Systems, Computer Vision, Brain Computer Interface, Intelligent Transportation System, and Internet of Things. She serves as Associate Editor for ACM Transactions on Sensor Networks, Knowledge-Based Systems, ACM Transactions on Recommender Systems, IEEE Transactions on Artificial Intelligence, and Editorial Board for Nature Scientific Reports.
报告摘要：
Reinforcement learning is an effective method for modeling dynamic user interests in recommender systems and has recently attracted increasing attention. However, it suffers from a drawback: low data efficiency due to its interactive nature. Training reinforcement learning-based recommender systems require expensive online interactions to gather sufficient trajectories, which are essential for the agents to learn user preferences. Recent strides in offline reinforcement learning present a new perspective. Offline reinforcement learning empowers agents to glean insights from offline datasets and deploy learned policies in online settings. I will briefly review existing literature on offline reinforcement learning for RecSys in this presentation and introduce our recent progress.
欢迎广大师生踊跃参加！
软件学院
2024年6月4日