CNN-Assignment-2025

NTCU113-2 ｜ Machine Learning ｜ 賴冠州教授

Submission Requirements

• Submit your work as a Jupyter Notebook (.ipynb).
• File name format: ClassNumber_CNN_Assignment.ipynb (e.g., ACS109145_CNN_Assignment.ipynb).
• Ensure the notebook includes all code, visualizations, and a report section answering Task 5.
• Upload to this repository via a pull request (PR).

Steps

1. Fork this repo.
2. Do Submission Requirements base on cnn_assignment.ipynb in colab.
3. attach your file(.ipynb) to your github repo
4. Commit and Create PR.

5 Key Tasks to Complete:

1. Task 1: Model Architecture Enhancement

   • Modify the CNN model structure
   • Must include model = models.Sequential and Conv2D layers
   • 要求：修改 CNN 模型架構
   • 檢查點：模型必須包含至少一個 model = models.Sequential and Conv2D layers
   • 提示：可以調整卷積層數量、濾波器數量、核大小等

2. Task 2: Hyperparameter Optimization

   • Implement model.compile and model.fit
   • Specify optimizer (SGD/RMSprop/Adam)
   • 要求：修改模型編譯時的超參數
   • 檢查點：必須指定優化器（如 SGD、RMSprop、Adam）
   • 提示：可以調整學習率、損失函數、優化器類型等

3. Task 3: Data Augmentation

   • Add ImageDataGenerator with augmentation parameters
   • Include: rotation_range, width_shift_range, height_shift_range, horizontal_flip
   • 要求：實現數據增強技術
   • 檢查點：使用 ImageDataGenerator 並包含增強參數
   • 提示：可以使用旋轉、平移、翻轉等增強技術

4. Task 4: Visualization

   • Create plots using plt.plot, plt.subplot, or plt.imshow
   • Generate predictions variable for model predictions
   • 要求：添加可視化功能
   • 檢查點：
   • 包含繪圖代碼（plt.plot、plt.imshow 或 plt.subplot）
   • 必須包含模型預測代碼
   • 提示：可以可視化訓練曲線、預測結果、混淆矩陣等

5. Task 5: Report Section

   • Add Markdown cell with heading containing "Task 5:", "Report", or "Conclusion"
   • Include meaningful analysis (more than 3 non-empty lines)
   • 要求：撰寫實驗報告
   • 檢查點：
   • 添加 Markdown 單元格，標題包含 "# Task 5:"、"# Report" 或 "# Conclusion"
   • 內容超過3行有意義的文字
   • 提示：描述實驗過程、結果分析、改進建議等

Autograding

• File name format validation
• Code execution without errors
• Presence of required code components
• Visualization outputs
• Report section completeness