supervised learning
Supervised learning is how a machine learns by example, with an answer key in hand. You hand it thousands of examples where each input is already paired with the right output — a photo labeled "cat," an email marked "spam" — and it gradually figures out the pattern that connects the two. It's like a student studying flashcards: question on the front, answer on the back, practiced until the rules sink in.
The whole point is generalization. Once trained, the machine should handle inputs it has never seen — a brand-new photo, tomorrow's email — and still give the right answer. The tasks come in two flavors: classification, where the output is a category (cat or dog, spam or not), and regression, where the output is a number (a house's price, tomorrow's temperature).
The catch — and it's a big one — is the labels. Every example must be tagged with its correct answer, usually by humans, which is slow and costly. And the machine only learns what the labels teach it: feed it biased or sloppy examples, and it will faithfully reproduce those flaws. "Supervised" doesn't mean someone watches over the machine as it works; it means a teacher supplied the answers during training.
Show it 10,000 photos each tagged "cat" or "dog"; afterward it labels a new photo on its own.
Classification: the labels teach it; then it generalizes to the unseen.
The name dates to the early decades of machine learning: the labeled examples act as a "supervisor" or teacher correcting the model's guesses. Its counterpart, unsupervised learning, gets no answer key and must find structure in the data on its own.