Deep learning dynamics from physical principles We propose an experimental approach to the isochronism problem using neural differential equations.