Principal Component Analysis
Core route · data intuition
Rotate your attention, not just your axes.
This explainer turns dimensionality reduction into something you can drag. Start in 2D and move points until the first principal component feels inevitable, then rotate the 3D cloud, and only afterwards jump into the 17D UK food example where the meaning comes from the separation you have already learned to see.
At a glance
PCA is often taught as a formula. This turns it into a physical rotation problem, making the connection between variance, projection, and dimensionality drop visceral.
Start in 2D and physically move the points until PC1 feels like the obvious axis of variation. Then rotate the 3D cloud and only after that open the 17D food dataset, where patterns you have already trained your eye to see become meaningful.
About 12 to 18 minutes. Intermediate. Best after one simpler model-intuition note so the geometry feels like an upgrade instead of a jump.
Embed
Reading path
- Open the live interactive: https://kohnnn.github.io/interactive-explanation/principal-component-analysis/
- Continue through Train-Test-and-Validation-Sets when you want a workflow-first companion to the geometry
- Continue via Interactive or Visual Notes