Predictive Modeling: Regression Fit Lab

Predictive modeling turns past observations into forecasts by fitting a mathematical relationship between predictors and an outcome, then checking whether that relationship generalizes to data it never saw during training. This simulator visualizes that whole loop: a polynomial regression surface is fit to a noisy 3D dataset by live gradient descent, with separate training and test point clouds so you can watch the classic underfitting-to-overfitting trade-off play out as model complexity, data noise and learning rate change, tracked in real time through train/test MSE and R².