[akmoleksandr]

ML is either taught via fit() and predict() without explaining what's happening inside, or with a bunch of university-level math. I wanted to fix that, so I ended up turning the process into a full book.

Each algorithm follows the same pattern:

- Plain-English intuition - Math formalization - NumPy implementation from scratch - Validate against Sklearn/PyTorch - Practical tips on when and why to use it

It covers Linear & Logistic Regression, Regularizations, Naive Bayes, KNN, Decision Trees, Random Forest, XGBoost and Neural Networks. Figuring out a simple way to implement and break down XGBoost was the toughest part but very much worth it.

It assumes basic Python and high-school math only.

GitHub: https://github.com/ml-from-scratch-book/code Book: https://a.co/d/0fmhuLbH

– cheers, alex