Machine learning · interactive lab
Decision Boundary
A classifier built from one rule: ask the nearest examples. Inspect its votes, change the neighborhood, and add a point to reshape the boundary.
Inspect the neighborhood
Click the plot or use its arrow keys to move the crosshair. Lines connect the query to its voting neighbors.
● Cyan = class A · ■ Magenta = class B. Stronger shading means more neighbors agree; it is not a calibrated probability.
The vote
Class A
2 votes for A · 1 votes for B
Query: (0.00, 0.00)
Change the evidence
Place the crosshair, choose a class, then add an example. Try planting class B inside a cyan region.
40 / 100 examples
Does it generalize?
100.0%
Leave-one-out accuracy: each example is classified using only the other points. This avoids the perfect training score k = 1 gets by looking up itself. It describes these synthetic examples, not performance on unseen real-world data.
Euclidean distance on equally scaled axes. Fixed synthetic datasets; no model downloads or training service.