Supervised Learning Tutorial

This course picks up where the basics stop. It is for people who already know what a training set and a test set are (if not, start with Machine Learning Fundamentals) and want to understand how supervised models really learn, how to choose between them fairly, and how results get faked by accident. You need Python and a little comfort with averages and slopes; no calculus beyond "the gradient points uphill" is assumed.

You will build linear regression with gradient descent and watch the loss fall, train a logistic regression and a small decision tree from scratch, use k-fold cross-validation to choose a model's complexity, and then confront the traps that make a model look better than it is: class imbalance, the wrong metric, and data leakage. Each trap is reproduced in code so you can see the symptom instead of taking it on trust.

The four lessons move from the mechanics of learning (regression, classification) to the discipline of judging (selection, then failure modes). Everything runs in plain Python with no dependencies, and every output shown was produced by running the code. Libraries like scikit-learn wrap the same ideas, so what you learn here transfers directly.