// the take
This is one professor's personal lecture archive — 2000+ slides of machine learning, probability, and deep learning theory covering everything from PAC-Bayes bounds to diffusion models and Neural ODEs, built up and revised over roughly a decade of teaching. It's for people who want rigorous, math-heavy derivations of ML theory topics rather than code — grad students, researchers, or anyone brushing up on a specific technique. There's no software to install or run here; it's PDFs, a few PPTX decks, and demo scripts referenced inline from a sister repo.
The math is actually derived, not hand-waved — variational inference, Dirichlet process mixtures, policy gradient theorems, and similar topics get full derivations rather than summary slides. Coverage keeps pace with current research: recent additions cover Deepseek's multi-head latent attention, KV caching, and flow matching alongside decade-old foundational notes on EM and MCMC. Most topics link a runnable demo (Matlab scripts in roboticcam/matlab_demos, a handful of Jupyter notebooks) and often a recorded lecture, so the derivations aren't left purely abstract.
It's a flat list of PDF and PPTX links in a long README, not a structured docs site — no search, no cross-referencing between related notes, no per-topic permalinks, so finding a specific derivation means scanning the whole page. Quality and notation are inconsistent across the collection: several notes are explicitly marked as 2015-era drafts awaiting a rewrite that hasn't happened yet. Demos live in a separate repo (matlab_demos) that can drift out of sync with the notes referencing it, and some material is legacy PowerPoint that's not really diffable or versionable. There's nothing to integrate into a project — this is reading material, not a library, so there's no programmatic way to consume any of it.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →