This project implements a from-scratch library for automated dynamical system recovery from (noisy) data based on SINDy. The algorithm is robust to the type of dynamics---even for chaotic systems and systems with bifurcations---and robust to varrying levels of noise. In fact, the algorithm is robust to noise applied pre-numerical differentiation, a result not shown in the original paper. As a further extension to the original paper, I present a set of techniques for quantifying the accuracy of the retrieved systems.
In addition to the code, see the paper I wrote for an overview of the background, methods, and results from the method. Further, I gave a presentation to my Dynamical Systems class overviewing my results.