trekhleb/homemade-machine-learning

Homemade Machine Learning is an educational repository that implements popular machine learning algorithms from scratch in Python, with interactive notebooks and mathematical explanations alongside each one. Rather than calling a single line of code from a library, every algorithm is written out step by step so you can see exactly how it works. The algorithms covered include supervised learning techniques like linear regression (predicting a number, such as a country's happiness score) and logistic regression (categorizing inputs, such as recognizing handwritten digits). It also covers unsupervised learning like K-means clustering (grouping data by similarity without labels) and anomaly detection (spotting unusual data points). Neural networks are covered too. Each algorithm has three components: a plain-English explanation of the mathematics, the Python source code, and an interactive Jupyter Notebook demo where you can experiment with the data and see the output in your browser in real time. This is aimed at people learning machine learning who want to understand how algorithms work at a conceptual and mathematical level, not just how to call a library function. The implementation is intentionally simple and not designed for production use. The tech stack is Python, with Jupyter Notebooks as the interactive environment.