Imitating Human Go Players via Vision Transformer

Developing AI algorithms for the game of Go has long been a challenging task. While tools such as AlphaGo have revolutionized gameplay, their focus on maximizing win rates often leads to moves that are incomprehensible to human players, limiting their utility as training aids. This work introduces a...

Full description

Saved in:
Bibliographic Details
Main Authors: Yu-Heng Hsieh, Chen-Chun Kao, Shyan-Ming Yuan
Format: Article
Language:English
Published: MDPI AG 2025-01-01
Series:Algorithms
Subjects:
Online Access:https://www.mdpi.com/1999-4893/18/2/61
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850081675219828736
author Yu-Heng Hsieh
Chen-Chun Kao
Shyan-Ming Yuan
author_facet Yu-Heng Hsieh
Chen-Chun Kao
Shyan-Ming Yuan
author_sort Yu-Heng Hsieh
collection DOAJ
description Developing AI algorithms for the game of Go has long been a challenging task. While tools such as AlphaGo have revolutionized gameplay, their focus on maximizing win rates often leads to moves that are incomprehensible to human players, limiting their utility as training aids. This work introduces a novel approach to bridge this gap by leveraging a Vision Transformer (ViT) to develop an AI model that achieves professional-level play while mimicking human decision-making. Using a dataset from the KGS Go server, our ViT-based model achieves 51.49% accuracy in predicting expert moves with a simple feature set. Comparative analysis against CNN-based models highlights the ViT’s superior performance in capturing patterns and replicating expert strategies. These findings establish ViTs as promising tools for enhancing Go training by aligning AI strategies with human intuition.
format Article
id doaj-art-9744850f92e643e98f51ebcd39bbfc94
institution DOAJ
issn 1999-4893
language English
publishDate 2025-01-01
publisher MDPI AG
record_format Article
series Algorithms
spelling doaj-art-9744850f92e643e98f51ebcd39bbfc942025-08-20T02:44:40ZengMDPI AGAlgorithms1999-48932025-01-011826110.3390/a18020061Imitating Human Go Players via Vision TransformerYu-Heng Hsieh0Chen-Chun Kao1Shyan-Ming Yuan2Department of Computer Science, National Yang Ming Chiao Tung University, Hsinchu 30010, TaiwanDepartment of Computer Science, National Yang Ming Chiao Tung University, Hsinchu 30010, TaiwanDepartment of Computer Science, National Yang Ming Chiao Tung University, Hsinchu 30010, TaiwanDeveloping AI algorithms for the game of Go has long been a challenging task. While tools such as AlphaGo have revolutionized gameplay, their focus on maximizing win rates often leads to moves that are incomprehensible to human players, limiting their utility as training aids. This work introduces a novel approach to bridge this gap by leveraging a Vision Transformer (ViT) to develop an AI model that achieves professional-level play while mimicking human decision-making. Using a dataset from the KGS Go server, our ViT-based model achieves 51.49% accuracy in predicting expert moves with a simple feature set. Comparative analysis against CNN-based models highlights the ViT’s superior performance in capturing patterns and replicating expert strategies. These findings establish ViTs as promising tools for enhancing Go training by aligning AI strategies with human intuition.https://www.mdpi.com/1999-4893/18/2/61Godeep learningVision Transformer
spellingShingle Yu-Heng Hsieh
Chen-Chun Kao
Shyan-Ming Yuan
Imitating Human Go Players via Vision Transformer
Algorithms
Go
deep learning
Vision Transformer
title Imitating Human Go Players via Vision Transformer
title_full Imitating Human Go Players via Vision Transformer
title_fullStr Imitating Human Go Players via Vision Transformer
title_full_unstemmed Imitating Human Go Players via Vision Transformer
title_short Imitating Human Go Players via Vision Transformer
title_sort imitating human go players via vision transformer
topic Go
deep learning
Vision Transformer
url https://www.mdpi.com/1999-4893/18/2/61
work_keys_str_mv AT yuhenghsieh imitatinghumangoplayersviavisiontransformer
AT chenchunkao imitatinghumangoplayersviavisiontransformer
AT shyanmingyuan imitatinghumangoplayersviavisiontransformer