Discovering sparse interpretable dynamics from partial observations

Abstract

Nonlinear dynamical systems are ubiquitous in nature and play an essential role in science, from providing models for the weather forecast to describing the chaotic behavior of plasma in nuclear reactors. This paper introduces an artificial intelligence framework that can learn the correct equations of motion for nonlinear systems from incomplete data, and opens up the door to applying interpretable machine learning techniques on a wide range of applications in the field of nonlinear dynamics.