Reconstruction of observed mechanical motions with artificial intelligence tools

Abstract The goal of this paper is to determine the laws observed trajectories assuming that there a mechanical system in background and using these continue motion plausible way. are represented by neural networks with limited number parameters. training follows extreme learning machine idea. We for different levels embedding, thus we can represent not only equation but also symmetries kinds. In recursive numerical evolution system, require fulfillment all laws, within determined precision. way, successfully reconstruct both integrable chaotic motions, as demonstrate example gravity pendulum double pendulum.