Huber Robust Regression: Weights and Fitted Lines (2D)
A scatter plot with the true line, the least-squares line and the Huber line converging by IRLS iterations, plus each point on the Huber weight curve. Same shared model and checks as the 3D page.
This simulator runs real genetic-programming symbolic regression against synthetic physical measurements — free-fall time versus drop height, pendulum period versus string length, or Kepler's third law relating orbital period to radius. A population of randomly built expression trees is scored by RMSE against noisy data, then evolved generation after generation through tournament selection, subtree crossover, and mutation. Every generation's best-fit curve is rendered at the front of a 3D tunnel while older generations recede and dim behind it, turning dozens of generations of evolutionary search into a single readable corridor that visibly narrows onto the true law as the run progresses.
A scatter plot with the true line, the least-squares line and the Huber line converging by IRLS iterations, plus each point on the Huber weight curve. Same shared model and checks as the 3D page.
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install