cpe342-machine-learning-2026

Imported from Palapluem/cpe342-machine-learning-2026 (AGENTS.md). Install upstream with npx skills add Palapluem/cpe342-machine-learning-2026. Copyright stays with the author.

CPE 342 Machine Learning — Workspace Rules & Operational Guidelines

This document serves as the authoritative, comprehensive Workspace Rules and System Prompt for the CPE 342: Machine Learning repository at King Mongkut's University of Technology Thonburi (KMUTT). All contributors and AI coding assistants must strictly adhere to these rules.

1. Repository Identity & Authorship Policy

1.1 Repository Ownership & README.md

• Sole Repository Owner:
  • Wisit Suwannao (วิศิษฐ์ สุวรรณเนาว์) — Student ID: 67070501042
  • Department of Computer Engineering, Faculty of Engineering, KMUTT
• Rule for README.md:
  • In the ## 👤 Author section of README.md, list ONLY Wisit Suwannao (67070501042). Do not include group partners here, as this is the personal coursework repository of the repository owner.

1.2 Assignment Reports & Notebook Submissions

• Authorship by Assignment:
  • Assignments 1, 2, 3, 4: 2 Group Members
    • 67070501027 นัธทวัฒน์ ปริมสิริคุณาวุฒิ (Natthawat Primsirikunawut)
    • 67070501042 วิศิษฐ์ สุวรรณเนาว์ (Wisit Suwannao)
  • Assignment 5 (Deep Neural Networks): 3 Group Members
    • 67070501027 นัธทวัฒน์ ปริมสิริคุณาวุฒิ (Natthawat Primsirikunawut)
    • 67070501042 วิศิษฐ์ สุวรรณเนาว์ (Wisit Suwannao)
    • 67070501067 พลวริษฐ์ วัฒนเหมรัตน์ (Polwarit Watthanahemmarat)
  • Assignment 6 (Convolutional Neural Networks): 2 Group Members
    • 67070501027 นัธทวัฒน์ ปริมสิริคุณาวุฒิ (Natthawat Primsirikunawut)
    • 67070501042 วิศิษฐ์ สุวรรณเนาว์ (Wisit Suwannao)
• Official Submission Notebook File Naming:
  • Course convention: (First 2 digits of ID)_(Last 4 digits of Member 1)_(Last 4 digits of Member 2)[_Member 3].ipynb
  • Examples:
    • Assignment 5: 1027_1042_1067.ipynb
    • Assignment 6: 67_1027_1042.ipynb

2. Directory Hygiene & File Management Rules

2.1 The Zero-Script Rule for Python (*.py)

• STRICT PROHIBITION: Never save any .py script files directly in the repository root or inside any assignment directory (assignment/Assignment *).
• Allowed Location for Scripts: All helper scripts, migration tools, figure verification scripts, or LaTeX processors MUST be created and executed exclusively inside the assistant's temporary scratch directory: <appDataDir>/brain/<conversation-id>/scratch/
• Pre-Commit Verification: Verify zero .py files exist in the repository tree before staging (git status, find . -name "*.py" or Get-ChildItem -Recurse -Filter *.py). If any .py files exist in the repository tree, delete them immediately.
• Strict Dataset / Artifact Cleanliness: Never commit raw dataset folders (e.g. Cat_Dog_data/, __MACOSX/), uncompressed batches, or intermediate archive zip files to git.

2.2 Git Commit & Synchronization

• Always Keep Up-to-Date: Every set of edits must be tested, compiled, staged, committed, and pushed to origin/main: https://github.com/Palapluem/cpe342-machine-learning-2026.git
• Semantic Commit Messages: Use standard conventional commit format:
  • feat(assignment-X): ... for new features or sections
  • fix(assignment-X): ... for bug fixes, alignment corrections, or layout adjustments
  • style(assignment-X): ... for formatting, typography, or spacing improvements
  • refactor(assignment-X): ... for restructuring or code optimization
  • docs(assignment-X): ... for documentation updates

3. LaTeX Academic Report Standards

3.1 Compilation Engine & Typography Setup

• Compiler: xelatex (run 2 passes to resolve all internal page references, counters, and citations).
• Language & Line Breaking:

  \usepackage[no-math]{fontspec}
  \usepackage{polyglossia}
  \setdefaultlanguage{thai}
  \setotherlanguage{english}
  \XeTeXlinebreaklocale "th"
  \XeTeXlinebreakskip = 0pt plus 1pt
  \sloppy

• Fonts:
  • Body Text: Sarabun with explicit font mapping:

    \setmainfont{Sarabun}[
        BoldFont       = Sarabun-Bold,
        ItalicFont     = Sarabun-Italic,
        BoldItalicFont = Sarabun-BoldItalic,
    ]

  • Monospace / Code Listings:

    \setmonofont{Consolas}[Scale=0.92]
    \newfontfamily\thaifonttt{Consolas}[Scale=0.92]

• Code Variables in Text: Never use raw \texttt{...} in descriptive text. Always use:

  \definecolor{codeaccent}{HTML}{0969DA}
  \newcommand{\codevar}[1]{\textcolor{codeaccent}{\textbf{#1}}}

• Listings Configuration:

  \lstdefinestyle{pycode}{
      language=Python,
      backgroundcolor=\color{codebg},
      basicstyle=\ttfamily\small,
      keywordstyle=\color{keyword}\bfseries,
      stringstyle=\color{string},
      commentstyle=\color{comment}\itshape,
      numberstyle=\tiny\color{comment},
      numbers=left,
      numbersep=8pt,
      frame=single,
      rulecolor=\color{codeframe},
      breaklines=true,
      breakatwhitespace=false,
      showstringspaces=false,
      tabsize=4,
      xleftmargin=16pt,
      xrightmargin=4pt,
      columns=flexible,
      keepspaces=true
  }
  
  \lstdefinestyle{output}{
      basicstyle=\ttfamily\small\color{output},
      backgroundcolor=\color{codebg},
      frame=single,
      rulecolor=\color{codeframe},
      numbers=none,
      xleftmargin=8pt,
      xrightmargin=4pt,
      breaklines=true,
      breakatwhitespace=false,
      columns=flexible,
      keepspaces=true
  }

3.2 Page Budgeting & Section Flow Rules

1. Section 5 (Q&A): Exactly 1 Page Per Question
   • Each individual question (e.g., Question 1, Question 2, Question 3, Question 4) must strictly occupy exactly one full page.
   • No question may spill over onto a second page. Adjust font sizing, item margins (itemsep), and table padding if needed to enforce this.
2. Seamless Section Transitions (Avoid Empty Pages):
   • Do not insert blind \clearpage before small subsections that leave $>50%$ of a page blank.
   • Let subsections and figures flow naturally. E.g., Section 4 follows Section 3.3 seamlessly; Section 7 concludes on the same page as Section 6.4.
3. Appendix Formatting (Assignment 5 Gold Standard):
   • Main Report Entry:

     \clearpage
     \appendix
     \raggedbottom
     \section*{Appendix: Full Jupyter Notebook Code \& Output}
     
     รายละเอียดต่อไปนี้คือโค้ด Python ที่ใช้แก้ปัญหาและวิเคราะห์แบบจำลอง [Model Domain] บนชุดข้อมูล [Dataset Name] พร้อมผลลัพธ์และกราฟทั้งหมด ซึ่งได้รันจริงจาก Jupyter Notebook (\texttt{<CourseNotebook>.ipynb} หรือ \texttt{<SubmissionNotebook>.ipynb})
     
     \input{notebook_appendix_X.tex}

   • Appendix File Structure (notebook_appendix_X.tex):
     • Starts with the Signature Blue Box:

       \begin{tcolorbox}[colback=blue!5!white,colframe=blue!75!black,halign=left,title=\textbf{Jupyter Notebook: <CourseNotebook>.ipynb\\ (ไฟล์ส่งงานหลัก: <SubmissionNotebook>.ipynb)}]
       โค้ด ผลลัพธ์ และการพล็อตภาพทั้งหมดด้านล่างเป็นการรันจริงจาก Jupyter Notebook
       \end{tcolorbox}

     • No Cell 0 Markdown Card: Jump immediately after the blue box into [In 1].
     • Figure Sizing & Captions in Appendix: Standardize all figures to width=0.96\linewidth. Always precede or accompany each plot output with its corresponding markdown explanation card (using tcolorbox with halign=left, before upper={\sloppy\raggedright}).
     • Strict Byte-for-Byte Accuracy (Zero-Tolerance): Every code [In] listing and plaintext [Out] listing in the LaTeX appendix must match the notebook cells byte-for-byte, including trailing newlines and exact strings.
     • Unconditional Figure Generation: In the notebook, all figures must be generated unconditionally using plt.savefig(...) without any if not Path(...).exists() caching bypasses or synthetic fallbacks.
     • Overfull / Underfull Zero-Warning Rule: Use \sloppy and \raggedright inside boxes containing code tokens or long URLs to guarantee zero XeLaTeX overfull \hbox warnings.

4. Jupyter Notebook Standards

4.1 Cell Structure & Quality

• Cell 0: Clean markdown header containing Course Title, Assignment Name, Authors with Student IDs, and Instructor (Dr. Boonyarit Changaival).
• No Trailing Blank Lines: Every code cell and markdown cell must be trimmed so there is zero trailing empty line at the bottom.
• Strict Reproducibility:

  import random, numpy as np, tensorflow as tf
  SEED = 42
  random.seed(SEED)
  np.random.seed(SEED)
  tf.random.set_seed(SEED)

• Sequential Execution Counts: Execution counts must strictly increment $1, 2, 3, \dots, N$ without gaps or missing outputs.
• Embedded Visualizations: All generated plots must be displayed inline and saved to vector .pdf files.

4.2 Matplotlib Chart Styling

• All figures must use the Sarabun font family:

  import matplotlib.pyplot as plt
  plt.rcParams['font.family'] = 'Sarabun'
  plt.rcParams['font.sans-serif'] = ['Sarabun', 'TH Sarabun New', 'DejaVu Sans']

• Use publication-grade DPI (150–300), clean academic color palettes (#1E3A8A, #2563EB, #16A34A, #DC2626), transparent gridlines (alpha=0.3), and bilingual/English academic annotations.

5. Course Domain & Assignment Roadmap