One of the three major forms of ML. Goal is to teach a model to reliably convert inputs into associated correct outputs using labeled data (input features paired with correct output labels). Analogous to learning from a teacher who provides correct answers.