Inference of dynamical systems evolution laws from raw data
Résumé
The advance of machine learning technology allows one to obtain useful information about the behavior of physical systems from raw datasets. In this context, state-of-art statistical techniques may be employed to discover the evolution laws of dynamical systems. This talk discusses the use of data-driven machine learning approach for inference of the differential equations that govern the behavior of a given dynamical system, for which sparse time-series are available. The methodology is based on library of mathematical functions and uses L1- regularized regression.