A Machine Learning paradigm where a model is trained on a labeled dataset — examples with known correct answers — so it can learn to make predictions on new, unseen data.
In Depth
In Supervised Learning, a model is trained on a dataset where each input example is paired with a known correct output label. The algorithm iteratively adjusts its internal parameters to minimize the difference between its predictions and the true labels. Once trained, the model can generalize — applying what it learned to make predictions on new data it has never seen.
Supervised Learning covers two fundamental problem types. Classification asks 'which category does this belong to?' — spam or not spam, tumor or benign, cat or dog. Regression asks 'what is the numerical value?' — predicting house prices, stock returns, or patient survival rates. Both rely on the same core principle: minimize prediction error on labeled examples to learn a generalizable function.
The critical challenge in Supervised Learning is acquiring enough high-quality labeled data. Labeling is expensive and time-consuming — medical images must be annotated by clinicians, legal documents by lawyers. Active Learning, Transfer Learning, and Semi-Supervised Learning are strategies to reduce this dependency. The alternative — Unsupervised Learning — avoids the labeling problem entirely, at the cost of less explicit signal.
Supervised Learning is the most widely used ML paradigm because it produces the most predictable, controllable results — provided you have enough labeled training examples of sufficient quality.