Education, Science, Technology, Innovation and Life
Open Access
Sign In

Application of Deep Learning Technology in Computer Go

Download as PDF

DOI: 10.23977/cii2019.70

Author(s)

Min Liu, Tiejun Feng

Corresponding Author

Min Liu

ABSTRACT

The objective of this study is to analyze the relevant deep learning techniques used by Alpha Go, understand the application of convolutional neural network algorithm in computer Go, and train a value neural network, which can evaluate the chess board of Go and meet the requirements of valuation in computer Go program. In this study, the relationship between deep learning and computer go was analyzed. Based on the convolutional neural network in deep learning, the core network of Alpha Go, that is, value neural network, was elaborated, and the related algorithms of value neural network were optimized to avoid the problem of over-fitting in value network training. It was found that deep learning, as a technology developed by traditional neural network, can perceive the information of computer go board and possess strong understanding and decision-making ability when solving the problem of computer go. In the process of value neural network training, some optimization algorithms were proposed, which can improve the computational efficiency and reduce the storage space at the same time. Therefore, the combination of artificial intelligence based on deep learning technology and computer go can improve the level of professional go players to a certain extent.

KEYWORDS

Deep learning, Computer go, Convolutional neural network, Artificial intelligence

All published work is licensed under a Creative Commons Attribution 4.0 International License.

Copyright © 2016 - 2031 Clausius Scientific Press Inc. All Rights Reserved.