Python notebooks that redraw the figures in Murphy’s ML books

What it does

This repository holds Python 3 notebooks that reproduce the figures in Kevin Murphy’s Probabilistic Machine Learning: An Introduction and Probabilistic Machine Learning: Advanced Topics. It relies on the standard scientific Python stack—numpy, scipy, matplotlib, sklearn—while pulling in JAX for the advanced material and occasionally TensorFlow 2 or PyTorch when the figure demands it. Shared utilities live in a separate probml-utils package to keep the notebooks focused on the math.

The interesting bit

Textbook companion code usually marries one framework and calls it a day. This repo treats JAX, TensorFlow, and PyTorch as situational tools, picking whichever fits the pedagogical point—a refreshing bit of pragmatism in a field prone to tribal loyalty.

Key highlights

• Covers two full volumes, from introductory Bayes to advanced topics, with notebooks sorted by chapter.
• Designed to run in Colab with free GPU/TPU access, or locally if you bring your own hardware.
• Publishes live dashboards tracking the reproducibility status of figures in both books.
• Entered maintenance mode in September 2022, so the collection is complete but largely frozen.
• Includes Google Summer of Code contributions from 2021 and 2022.

Caveats

• The codebase is in maintenance mode as of September 2022, so expect fixes but not new chapters.
• The README warns that the Git history is “very large”; a shallow clone is effectively mandatory.
• This is companion code, not a standalone library—expect figure scripts, not importable abstractions.

Verdict

If you are working through Murphy’s books and want to tweak priors or re-render plots, this is essential. If you need a general-purpose probabilistic programming toolkit, look elsewhere.