The map

Three styles, set by your data.

Supervised: learn from labelled examples. Unsupervised: find structure in unlabelled data. Reinforcement: learn by trial, error, and reward. Which one you use depends entirely on what data you have.