What it does

This repo is a personal learning archive: ~30 Jupyter notebooks covering the standard machine learning curriculum—regression, classification, clustering, NLP, neural networks, reinforcement learning, and the evaluation metrics that go with them. Each topic typically includes theory, from-scratch implementations, and scikit-learn equivalents. The author compiled them while working through Andrew Ng’s Coursera course, several Udemy courses, and the scikit-learn cookbook.

The interesting bit

Most “awesome-ml” lists are link rot. Here the notebooks are self-contained, hosted on nbviewer, and organized by topic rather than by source course. The regression section even includes the often-skipped bits: assumptions checking, dummy variable traps, backward elimination, and robust regression with Theil-Sen. Someone actually did the homework.

Key highlights

• From-scratch linear regression with gradient descent, plus the scikit-learn version for comparison
• Feature preprocessing as its own section—imputation, encoding, scaling, and text vectorization
• Clustering applications beyond toy datasets: image quantization and outlier detection
• Backpropagation derived and then implemented in plain Python
• Coverage of less-glamorous topics: association rule mining (Apriori, Eclat), cross-validation types, silhouette distance

Caveats

• No installation instructions, requirements.txt, or tested environment; you’re on your own for dependencies
• “Compiled while learning” means quality may vary; some notebooks are theory-only, others are code-only
• Last substantial update unclear; several nbviewer links could stale-date

Verdict

Good for someone who wants a second perspective after a video course, or needs a quick refresher on a specific algorithm’s mechanics. Not a substitute for a textbook or a maintained library. If you’re past the “implement backprop by hand” stage, this is mostly nostalgia.