Sparse Nonlinear Dynamics Models with SINDy, Part 5: The Optimization Algorithms

This video discusses the various machine learning optimization schemes that may be used for the Sparse Identification of Nonlinear Dynamics (SINDy) algorithm. We discuss the LASSO sparse regression, sequential thresholded least squares (STLS), and the sparse relaxed regularized regression (SR3) algorithms. We also discuss how to enforce partially known physics, such as energy conservation in incompressible fluid flows, by constraining the sparse least squares regression. Citable link for this video at: https://doi.org/10.52843/cassyni.f1bn05 Original SINDy paper: https://www.pnas.org/content/113/15/3932 @eigensteve on Twitter eigensteve.com databookuw.com Video to enforce stability: https://www.youtube.com/watch?v=ysg_8yHiO5k Paper to enforce stability: https://journals.aps.org/prfluids/abstract/10.1103/PhysRevFluids.6.094401 This video was produced at the University of Washington %%% CHAPTERS %%% 0:00 Introduction & Recap 1:56 Parsimonious Modeling with SINDy 9:25 SR3: Sparse Relaxed Regularized Regression 11:27 Constrained SINDy: Enforcing Energy Conservation 16:38 SINDy for Abrupt System Changes