With the further development of machine learning methods, algorithms are in the face of a more complicated environment than ever. The massive amount of data has brought severe pressure to present computing and decision-making strategies. In order to augment the performance of artificial intelligence in multi-agent interacting scenarios, researchers have adopted game theories to machine learning strategies. The article introduces the application of game theory from three aspects: the theoretical basis, the application fields, and the future prospects. The main purpose of the article is to summarize the current research of applying game theory to reinforcement learning algorithms and how the theory improves the performance of the witch. It is hoped that the article can provide a guide for interested researchers, and help push forward the achievements of the field. The research on multi-game theories not only improves the ability of logistic game related algorithms, but also positively pushes the development of very advanced artificial intelligence further.
Chang Yuan (Thu,) studied this question.