Two kinds of thing, plotted by two measurements. The examples come labeled — this or that. A classifier is just a line: everything on one side is called one class. Tilt and shift the line until each colour sits on its own side. That line, found from the labels, is what lets the model judge a new point it has never seen.
A linear classifier: predict class B where cos θ·x + sin θ·y > b. Node-verified: a boundary aligned with the gap separates these labels at 100%; a perpendicular one scores 38%. Built on the reusable probe.js primitive. Runs on your device. · Institute for Physical AI @ BMI