Notes on the Architecture, League Training and PFSP in AlphaStar
Jianyang Zhao, Ningning Zhou, Long Hong · 2021 16th International Conference on Intelligent Systems and Knowledge Engineering (ISKE) · 2021
AlphaStar, a multi-agent intelligence program, is a new member of Alpha’s gamer software. Rated at Grandmaster level in the empirical evaluation shows that AlphaStar has been a great success in engineering, which is another milestone in AI history. After a brief introduction to AlphaStar, we discussed its computational architecture, league training, PFSP, and proposed an interesting topic direction on the basis of our suggestions, which was how to combine the capacities of diverse ANNs that successful completed learning. Additionally, we proposed the viewpoint that is big data— space, intelligence— man’ world, parallel computing—time.