Notebooks — overview

This section targets ML-minded readers — same models as the main text, in Python, editable hands-on.

How to open

1. Read here — static pages (KaTeX + embedded plots).
2. Live Jupyter — after deploy open /lab/: JupyterLab in-browser (Pyodide/WebAssembly; first visit ~30 MB). Pick notebooks from repo folder study-guide/notebooks/ (bundled into JupyterLite builds).
3. Local build — pnpm build:with-lab emits JupyterLite under dist/lab/ (needs Python 3 with jupyterlite-core and jupyterlite-pyodide-kernel; script installs them).

Want Colab? Copy code from these pages into your own notebook.

Notebook roadmap

1. BKT from scratch — numpy model in ~30 lines; digits match TS exactly. ✓ ready
2. Parameter sensitivity — what breaks when $P(G)=0.5$ or $P(T)=0$. ✓ ready
3. EM fitting — Baum–Welch parameter recovery on synthetic data. Answers “where do parameters come from?” ✓ ready
4. IRT vs BKT — comparison with Item Response Theory — when each wins. ✓ ready
5. Class simulation — 22 students × 8 skills × 50 lessons — pitch charts. ✓ ready