Deep Reinforcement Learning for Game AI: A Case Study in StarCraft II

Publisher：闻天明Release Time：2021-12-28Number of visits：213

Speaker:     Dr. Junxiao Song, inspir.ai (启元世界)
Time:          9:00-10:00 , Dec.29
Host:           Prof. Ziping Zhao
Tencent Meeting:   389-191-380
Abstract:
AlphaStar, the AI that reaches GrandMaster level in StarCraft II, is a remarkable milestone demonstrating what deep reinforcement learning can achieve in complex Real-Time Strategy (RTS) games. However, the complexities of the game, algorithms and systems, and especially the tremendous amount of computation needed are big obstacles for the community to conduct further research in this direction. We propose a deep reinforcement learning agent, StarCraft Commander (SCC). With order of magnitude less computation, it demonstrates top human performance defeating GrandMaster players in test matches and top professional players in a live event. Moreover, it shows strong robustness to various human strategies and discovers novel strategies unseen from human players. In this talk, we will share the key insights and optimizations on efficient imitation learning and reinforcement learning for StarCraft II full game.

Bio:
Dr. Junxiao Song is a Research Scientist with inspir.ai (启元世界), in charge of theoretical research and system development related to reinforcement learning and decision intelligence. The StarCraft Agent he has developed with his team reaches the level of professional players. Relevant research results are published at NeurIPS and ICML. Before joining inspir.ai, Dr. Song worked at NetEase Games as a Senior Data Mining Researcher, responsible for the research team in personalized recommendation systems and Game AI technology based on deep reinforcement learning. Dr. Junxiao Song obtained his Bachelor degree from Zhejiang University in 2011 and was awarded the Distinguished Zhu Kezhen Scholarship, the highest honor for students of Zhejiang University. In 2015, he obtained his PhD degree from the Hong Kong University of Science and Technology. Dr. Song Junxiao has published papers in well-known international journals and conferences in the field of machine learning, signal processing, and optimization (such as IEEE Transactions on Signal Processing, NeurIPS, ICML, etc.), and co-authored one book on reinforcement learning.