My research focuses on proving that adding governing equations as the constraint on the training of neural networks results in the network generalizing better and requiring less data.