A machine learns from data — no explicit rules needed. Watch patterns emerge, boundaries form, and clusters appear live in 3D.
Traditional programming: you write rules, data goes in, answers come out. Machine learning flips this — you give data and answers, and the computer writes its own rules. It finds structure, boundaries, and patterns at a scale no human can match by hand.
Switch between modes to see how supervised vs unsupervised learning look different in the same data space.
Every ML algorithm belongs to one of these three families. The data and the task dictate which to reach for.
You provide input–output pairs and the model learns the mapping. At inference, it predicts the output for inputs it has never seen.
Examples: spam detection, house price prediction, image classification, fraud detection.
No labels — the model finds hidden structure in raw data on its own. It discovers clusters, patterns, and compressed representations.
Examples: customer segmentation, anomaly detection, topic modeling, dimensionality reduction.
An agent takes actions in an environment and receives rewards or penalties. It learns through trial and error to maximise long-run reward.
Examples: game playing (AlphaGo), robot locomotion, self-driving cars, recommendation systems.
Five terms you will encounter in every paper, tutorial, and codebase.
The input variables — the measurable properties of each data point. Also called attributes or predictors.
The output the model must predict. A class name (classification) or a number (regression).
The learned function that maps features to predictions. It is a mathematical object, not a human.
The process of adjusting the model's internal parameters so it makes better predictions on known data.
How well the model performs on data it has never seen. The real goal — not just memorising the training set.