Hybrid Physics-Informed Neural Networks For Dynamical Systems