The Application of Reinforcement Learning-Alphago

The application of reinforcement learning which is used in the field of Go has important value. Because Go emphasizes both local attacks and defenses. Meanwhile, whether the situation is advantageous should be considered. Players need strong intuition and judgment. Simple mathematical models are difficult to describe the situation of the game reasonably. The article summarizes a series of algorithms from AlphaGo Fan to Alphazero based on the development history of algorithms. The article explains the principle of playing chess pieces in the AlphaGo series of algorithms. The article also focuses on the introduction of AlphaGo Master, AlphaGo Zero, and AlphaGo Zero three models from two aspects: Monte Carlo tree search and deep neural network. Afterwards, this article organized the dataset and evaluation criteria. Finally, this article summarizes the applications of AlphaGo in other fields and provides two ideas to improve the model. This article helps to understand the principles and basis for establishing the basic models of the AlphaGo series of artificial intelligence, and provides useful references for future research on related algorithms and models of the AlphaGo series. This article contributes to understanding the basic model of AlphaGo and provides useful assistance for future research.