01 — Explore
Fit the signal, not the noise
Click the plot to add points, drag them around, and see how the choice of features and penalty decides between a curve that generalizes and one that memorizes.
02 — Under the hood
What the penalty does
Standardized weights of each feature. Lasso sets some to exactly zero (flat ticks).
Every line is one weight. Move λ right and they shrink towards zero.
Training Cross-validated (± 1 SE). The gap is overfitting.