Sparse Identification of Dynamics: Thresholding a Library of Terms (2D)
Noisy samples of dx/dt against x with the true law and the sparse fit, beside bars comparing the true and recovered coefficient of each library term. Same shared model and checks as the 3D page.
Recover the equation behind noisy measurements of a rate of change. Build a library of candidate terms, fit them all by least squares, then repeatedly zero the small coefficients and refit the rest. Change the noise, the number of samples, the threshold and the true law, and watch the active terms, the coefficient error and the fitted curve. A Blender-built plotting easel with normal-equation, noise-free recovery and Monte Carlo standard-error checks built in.
Noisy samples of dx/dt against x with the true law and the sparse fit, beside bars comparing the true and recovered coefficient of each library term. Same shared model and checks as the 3D page.
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install