Machine Learning Notebooks (Offline)

Self-contained Jupyter Notebooks for classical ML, neural networks, and PyTorch implementations. All example datasets are stored in data/ (CSV/NPZ), so the notebooks run offline.

Structure

• classical_ml/ — Linear/Logistic Regression, SVM (NumPy)
• neural_networks/ — Single-layer NN, MLP (from scratch, NumPy)
• pytorch_versions/ — Single-layer NN (softmax), MLP, CNN, RNN (PyTorch)
• advanced_topics/ — Attention demo, Tiny Transformer
• data/ — Offline datasets (CSV/NPZ)

Usage

1. Create environment and install basics:

   pip install numpy pandas matplotlib torch

2. Launch Jupyter and open any notebook:

   jupyter notebook

3. Notebooks are designed to run cell-by-cell without external downloads.

Notes

• Plots use matplotlib; no style dependencies.
• Toy datasets are small and deterministic (fixed seeds) for reproducibility.
• Transformer notebook uses PyTorch's nn.Transformer on a tiny digits→words mapping.