// the take
A personal grab-bag of IPython notebooks from Olivier Grisel, a scikit-learn core maintainer, covering ML experiments from classic scikit-learn demos to GMMs, dask-distributed training, attention mechanisms, and tabular deep learning comparisons. It's for developers who already know ML and want to see how a sklearn insider pokes at edge cases, not a tutorial series for beginners.
Several notebooks demonstrate techniques that aren't well covered in official docs — variable importance with completely randomized trees, learning-to-rank data prep, semi-supervised extra trees, GMM model selection comparisons. The author's scikit-learn core involvement means the code reflects real internals knowledge, not surface-level API usage. An environment.yml and Binder badge let you run notebooks without local setup.
The README states outright these are 'mostly unfinished ML related experiments' — there's no guarantee any given notebook runs cleanly or reaches a real conclusion. No per-notebook writeup or ordering, so you're on your own figuring out which ones are worth your time versus half-finished scratch work. Several touch long-dead stack pieces (IPython.parallel, Numba Parakeet Cython) that won't reflect current best practice. No tests or CI, so there's no signal on whether notebooks still execute against current numpy/scikit-learn releases.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →