MODELING UNKNOWN DYNAMICAL SYSTEMS WITH HIDDEN PARAMETERS

We present a data-driven numerical approach for modeling unknown dynamical systems with missing/hidden parameters. The method is based on training deep neural network (DNN) model the system using its trajectory data. A key feature that contains parameters are completely hidden, in sense no information about available through either measurement data or our prior knowledge of system. demonstrate by DNN sufficient time history, resulting can accurately For new initial conditions associated new, and unknown, parameters, produce accurate predictions over longer time.