研究生: |
吳冠毅 Wu, Guan-Yi |
---|---|
論文名稱: |
圍棋詰棋教學系統之設計及研發 The Design and Development of a Teaching System for Tsumego |
指導教授: |
林順喜
Lin, Shun-Shii |
學位類別: |
碩士 Master |
系所名稱: |
資訊工程學系 Department of Computer Science and Information Engineering |
論文出版年: | 2020 |
畢業學年度: | 108 |
語文別: | 中文 |
論文頁數: | 59 |
中文關鍵詞: | 圍棋 、神經網路 、詰棋 、電腦圍棋教學 |
英文關鍵詞: | Go, Neural Network, Tsumego, Computer Go Teaching |
DOI URL: | http://doi.org/10.6345/NTNU202001481 |
論文種類: | 學術論文 |
相關次數: | 點閱:173 下載:0 |
分享至: |
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
圍棋是一項歷史悠久且對局複雜度較高的遊戲,近年來電腦對局遊戲領域發展逐漸成熟,人工智慧逐漸克服當初認為較複雜的賽局遊戲。然而,隨著棋類遊戲達到人類難以企及的程度時,人類也發現到無法理解電腦為何這樣下棋的問題。現有的圍棋軟體絕大部分僅能提供人類形勢判斷、勝率分析、變化圖等功能,於教學部分尚有一段需要努力的空間。
本研究針對電腦圍棋詰棋教學系統,研發一套方法,讓電腦學習如何判斷盤面落子原因及其目的性。我們先將人類知識透過人工註記的方式,將每個盤面的每手棋標記,透過卷積神經網路技術將人類知識做學習後,再替人類標記每手棋的名稱以及其目的為何。
在實作上,我們蒐集1500題詰棋題目,先分析並記錄每一局詰棋的每手棋之名稱、目的性,透過監督式學習訓練神經網路。最後可利用此系統將新的詰棋棋局的棋步內容,以 SGF (Smart Game Format) 檔案反饋予使用者,藉此達到圍棋教學的效果,期望在電腦圍棋教學上貢獻棉薄之力。
Go is a relatively complex game with a very long history. In recent years, the field of computer games has gradually matured, and artificial intelligence gradually overcomes many complex games. However, as the computer programs climbed at a level that was difficult for humans to reach, humans discovered that they cannot understand why the computer makes a move like this. Most of the existing Go software can only provide humans with functions such as situation judgment, win rate analysis, and variation diagram. In the teaching part, there is still much room for improvement.
This thesis aims at designing a teaching system for tsumego, and developing a set of methods for the computer to learn how to judge the reason and purpose of the stone placement. We first use human knowledge through manual annotation to label each move of each board, then let the system learn human knowledge using convolutional neural network trained by supervised learning, and hopefully the system can get the reason and purpose of each move for a new tsumego for humans.
For the implementation, we collect 1500 tsumego questions, analyze and record the reason and purpose of each move in each question. Then we train the neural network through supervised learning and finally the system can display the information of each move in a new tsumego question through the SGF (Smart Game Format). We expect to contribute our mite to the teaching field of computer Go.
[1] 維基百科,圍棋,https://zh.wikipedia.org/wiki/%E5%9B%B4%E6%A3%8B。
[2] 維基百科,詰棋,https://zh.wikipedia.org/wiki/%E8%A9%B0%E7%A2%81。
[3] 陳祖源,圍棋規則演變史,上海文化出版社,2007。
[4] 顏士淨,電腦圍棋程式 Jimmy 5.0 之設計與製作,國立臺灣大學資訊工程所博士論文,1999年。
[5] 李國弘,棋串攻殺系統的實作,國立東華大學資訊工程所碩士論文,2005年。
[6] 文亞成,電腦圍棋程式中棋形辨識處理之研究,淡江大學資訊工程所碩士論文,1989年。
[7] 李佑堂,應用 Patricia Tree 做為圍棋棋形辨識處理之研究,國立東華大學資訊工程所碩士論文,2004年。
[8] 張嘉麟,基於MongoDB建置電腦圍棋棋譜搜尋系統,東海大學資訊工程所碩士論文,2015年。
[9] Smart Game Format (SGF) website,https://www.red-bean.com/sgf/。
[10] AlphaGo,https://zh.wikipedia.org/wiki/AlphaGo。
[11] D. Silver, J. Schrittwieser, K. Simonyan, I. Antonoglou, A. Huang, A. Guez, T. Hubert, L. Baker, M. Lai, A. Bolton, Y. Chen, T. Lillicrap, F. Hui, L. Sifre, G. V. D. Driessche, T. Graepel, and D. Hassabis, “Mastering the game of Go without human knowledge,” Nature, vol. 550, no. 7676, pp. 354–359, 2017.
[12] 程冠倫,以深度學習解決圍棋死活問題,國立東華大學資訊工程所碩士論文,2017年。
[13] 湯曉鷗,陳玉琨,人工智能基礎,華東師範大學出版社,2018。
[14] 林大貴,TensorFlow + Keras 深度學習人工智慧實務應用,博碩出版社,2017。