HumachLearnLearn · explore · apply
Start learning
Working playground

Regression Playground

Fit multiple browser-side regressors to linear, nonlinear, noisy and heteroscedastic problems. Inspect both the fitted function and residual structure.

Lab concept guide

What to observe while you experiment

A regression model estimates a numeric response function from features. Linear, polynomial, KNN/tree/ensemble-style regressors respond differently to curvature, heteroscedasticity, outliers and regions without training support.

MechanismKeep the generated data fixed, train several regressors and compare fitted functions together with residual structure.
Failure modeSelecting the curve that visually passes closest to training points without checking validation error, residual pattern or extrapolation.
VerificationCalculate residuals for several points and compare a held-out metric across models under the same generated dataset and split.
Experiment deliberately
Generate linear, nonlinear and heteroscedastic cases. Predict which model assumptions will fail, then identify the failure in the fitted function and residuals.
Ready.