Ray, including Ray/RLLib, has been actually speeding up many reinforce-learning-based services in NetEase Games. It has greatly promoted the user experience of game players and the general profit of various game products as well. We deployed our first reinforcement learning based recommendation system application using ray RLLib. Reinforcement learning is a promising direction since the RL paradigm is inherently suitable for tackling multi-step decision-making problems, optimizing long-term user satisfaction directly, and exploring the combination spaces efficiently. As a way of giving back to the community, we open-sourced the RL4RS (Reinforcement Learning for Recommender Systems) dataset - a new resource fully collected from industrial applications to train and evaluate RL algorithms with special concerns on the reality gaps, contains two real-world datasets, data understanding tools, tuned simulation environments, related advanced RL baselines, batch RL baselines, and counterfactual policy evaluation algorithms. The RL4RS suit can be found at this repo (https://github.com/fuxiAIlab/RL4RS).
Runze Wu received the Ph.D. degree from University of Science and Technology of China. He is currently a Senior Researcher and the head of User Profiling Research Group in NetEase FUXI AI Lab, China. His research interests include user profiling, anomaly detection, causal inference, combinational optimization, deep learning, and various data mining and artificial intelligence applications across online games. He has published more than 25 papers in refereed journals and conference proceedings, such as ACM Transactions on Intelligent Systems and Technology, ACM SIGKDD, IJCAI, AAAI, TheWebConf, and ACM CIKM.
Come connect with the global community of thinkers and disruptors who are building and deploying the next generation of AI and ML applications.