Deep Reinforcement learning and its commercial applications
Speaker: Dr. Liang Zhang
Topic: Deep Reinforcement learning and its commercial applications
Time & Date: 14:00 -15:00, Tuesday,January 16, 2023 (Beijing time)
Zoom Meeting:
https://cuhk
Zoom Meeting ID:980 5873 5363
Password：123456
Abstract:
Since Google's AlphaGo outperformed the level of human in Go, reinforcement learning has proven its excellent ability in multiple game scenarios. At present, many research works continue to improve the ability of reinforcement learning algorithms to make them more efficient, faster and more stable, but the application of reinforcement learning in real commercial scenarios still faces great challenges. In this talk, we share the application research of reinforcement learning in two important commercial scenarios, namely recommender systems and game commercialization. In recommender systems, the characteristics of serialized recommendation by users and the pursuit of long-term benefits make reinforcement learning a suitable method on the recommender system. We solve the sequence decision-making problem faced by recommender systems from three aspects, namely Pair-wise reinforcement recommendation, Page-wise reinforcement recommendation and hierarchical reinforcement learning to improve both the click-through rate and the conversion rate. In the field of games, we introduce a new scenario for game commercialization, namely the AI-Bot matchmaking system. We propose an novel incremental system based on reinforcement learning, which is successfully deployed on the matchmaking system of a MOBA game.
Biography:
Dr. ZHANG Liang is an associate researcher in Peng Cheng Laboratory and selected as Shenzhen Peacock Program C talent. Dr ZHANG graduated from The Hong Kong Polytechnic University in 2016; from 2017 to 2022, he joined JD.com and Tencent, and published a number of reinforcement learning decision-making research papers such as KDD and SIGIR, which were successfully applied in commercial advertising in JD and The Honor of King in Tencent. He has published more than 20 papers with 1000+ Google citations. His research interests include reinforcement learning ant its applications, green buildings, and network economics.
