These notes represent my own study jottings and learning synthesis, not an authoritative or foolproof standard. If you find them helpful, feel free to use them. If you spot anything that could be clearer or have suggestions for additional concepts to include, issues and pull requests are warmly welcome.
Regression
Continuous numerical prediction and first-order optimization
03. Regression
Predicting continuous quantities, choosing a loss that matches the decision, and understanding how optimization actually moves.
Regression is the part of supervised learning that answers "how much?" rather than "which class?" That difference sounds small, but it changes the loss, the evaluation, and the way errors behave. A classification model that is 51% confident still makes a discrete choice; a regression model that is off by 10% produces a number that may be used directly in a price, a forecast, or a budget.
This chapter builds from the simplest continuous model to the optimization procedure that trains it. The goal is not just to recognize the equations, but to know what each term is doing and when the assumptions behind them break.