Ray/RLLib-Empowered Reinforcement Learning Based Recommender Systems in NetEase Game

Explore how Ray and Ray/RLLib are revolutionizing reinforcement learning-based recommender systems in NetEase Games through this 30-minute conference talk. Discover how these technologies have significantly enhanced user experiences and increased profits across various game products. Learn about the deployment of the first reinforcement learning-based recommendation system application using Ray RLLib, and understand why reinforcement learning is a promising direction for tackling multi-step decision-making problems, optimizing long-term user satisfaction, and efficiently exploring combination spaces. Gain insights into the open-sourced RL4RS (Reinforcement Learning for Recommender Systems) dataset, a valuable resource for training and evaluating RL algorithms with a focus on bridging reality gaps. Examine the contents of the RL4RS suite, including real-world datasets, data understanding tools, tuned simulation environments, advanced RL baselines, batch RL baselines, and counterfactual policy evaluation algorithms. Access the slide deck for a comprehensive overview of the talk's content and explore the potential of Ray and Ray/RLLib in advancing AI-driven recommender systems for the gaming industry.