eugene87222/intro-to-ai-group-project

AI of modified version of Othello/Reversi

Hi :))))

---

EXPERIENCE

Senior R&D Engineer @ Synopsys Inc.

• Hsinchu City, Taiwan
• Sep 2022 - Present

Part-time Worker @ National Center for High-Performance Computing

• Hsinchu City, Taiwan
• Jan 2018 - Present

• Design hundreds of web crawlers unaided with Python for retrieving textual, image, and video data from Internet forums and news websites, e.g., PTT, BBC, Google Images, and Flickr
• Conduct sentiment analysis to textual data by incorporating the Word2Vec model and traditional machine learning techniques
• Apply topic modeling on textual data using TF-iDF and classify the data by topics with the one-class support-vector machine
• Research into speech recognition model and speech-to-text cloud service provided by Google GCP and IBM Waston

Junior Engineer @ MiTAC Information Technology Corp.

• Neihu District, Taipei City, Taiwan
• Jul 2019 - Aug 2019

• Developed a registry editor GUI which allows users to manage registry handily on Windows operating system
• Collaborated with six colleagues to develop the software of ticket vending machine for Taiwan High Speed Rail
• Implemented the frontend and the backend of ticket vending software with Applied Windows Presentation Foundation (XAML, C#) and C++ respectively while communicating with each other via TCP socket

EDUCATION

National Yang Ming Chiao Tung University (NYCU)

• Sep 2020 - Sep 2022

• Master of Science in Multimedia Engineering
• GPA: 4.04/4.3
• Excellence Award on 2021 fall

National Chiao Tung University (NCTU)

• Sep 2016 - Jul 2020

• Bachelor of Science in Computer Science
• GPA: 3.76/4.3

intro-to-ai-group-project's Issues

td learning 和 searching agent 比較效果

目前上傳的版本 分別是和 all_3_black 對戰 10000場 和 20000 場的訓練的 weight
all_3_10000.p
all_3_20000.p

之前的 td_agent 版本練的效果是勝利不到25%
目前這個版本拿 20000 對抗 下面這個設定的 agent
black = SearchingAgent(True, 4.9 , 2.0, 1.0, 1.5, algo='pvs')
只持白棋 似乎有稍微能對抗一下 （大約50%）

還需要更多測試
all_2_tuple 的版本 也可以練一份看看效果
据 reference 提到的是 all_2_tuple 訓練速度結論來説比 all_3 優一些

TD learning 缺陷

目前 TD learning function 的 weight 依然會持續長歪
待處理

TD update 公式
s = s + alpha * ( (r + s') - s )
alpha = > learning rate (0~1)
r => reward of next state [-1, 1]
s' => next state
s => before state

棋子 統計狀況 討論

要針對 pattern 檢查可能有點困難

我目前只隔離在輸贏情況

遊戲過程選擇下棋位置的分佈

1. 黑棋贏
2. 黑棋輸
3. 白旗贏
4. 白旗輸
   概況:
   (a) 白旗完全迴避在綠色位置下棋
   (b) 白旗和黑棋佔邊次數很平均

---

結束，輸贏時擁有屬於黑白 擁有的 棋子位置分佈

同上（1. 黑棋贏
2. 黑棋輸
3. 白旗贏
4. 白旗輸）

概況:
(a) 白旗在贏得局面 經常爭取反轉中間的棋子

Progress

TODO

• get valid move (PR #3)
• 想 evaluate function
  • 權重 (commit 3d2a6a3)
    • 四條邊上棋子的數量 * 2.0
    • 棋子總數 * 1.5
    • valid moves 的數量 * 1.8
  • 權重可能還要修改
  • 想想還有沒有什麼東西可以加入 evaluation function
• minimax algorithm
  • 新增 prototype (commit 0bd4f7a)
  • first version，可以跟 AI 對局了 (commit bc77d36)
• alpha-beta pruning
  • 新增 prototype (commit 0bd4f7a)
  • first version，可以跟 AI 對局了 (commit bc77d36)
  • 看要不要增搜尋深度，目前是 100