Home▸Machine Learning & Neural Networks▸Classifier Threshold Sweep

3D Model Evaluation Simulator

Fly through a 3D point cloud of a classifier's score distribution as a decision plane auto-sweeps across it, tracing the ROC curve live in space with real, cumulative AUC computed as the sweep completes.

Machine Learning & Neural Networks3DEasy60 FPS📱 Mobile-adapted⇄ 2D version

The ROC curve is usually drawn flat and static — a line you're told to trust. This simulator builds it instead: a 3D field of classifier scores, one point per sample, with a decision plane that physically sweeps through the cloud from high scores to low. Every sample the plane has crossed counts toward the confusion matrix at that instant, and a live ROC trace grows on a wall in the same scene, point by point, as real trapezoidal AUC accumulates behind it. Orbit around the sweep, tune class separation and noise, and watch exactly how a threshold choice reshapes true and false positives in space before it ever becomes a single number on a curve.

⚙ Under the hood

Use a moving decision plane through a 3D point cloud of classifier scores to dynamically trace the ROC curve in real-time.

model-evaluationclassificationROC curve

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install