Classical models built from the math you already know
A thousand-pixel photo of a face wastes most of its numbers: shift a fraction of an inch and neighbouring pixels barely change, so most of that huge pixel count is redundancy, not information. Squeeze the same photo down to a well-chosen handful of numbers along the directions where it actually varies, and very little of what mattered about the face gets lost. Principal component analysis finds exactly those directions for any dataset, not just photos. It finds the best angle for flattening a high-dimensional cloud of points down onto fewer dimensions, keeping as much of the original spread as it can.
🔒 This is a Pro lesson — the interactive figure, worked examples, quiz and practice open with Pro access.
▶ PCA as a Model: Compression & Reconstruction