Emulating quantum dynamics with neural networks via knowledge distillation

We introduce an efficient training framework for constructing machine learning-based emulators and demonstrate its capability by artificial neural network to predict the time evolution of quantum wave packets propagating through a potential landscape. This approach is based on idea knowledge distillation uses elements curriculum learning. It works set simple, but rich-in-physics examples (a curriculum). These are used emulator learn general rules describing system (knowledge distillation). show that this capable learning dynamics from simple (wave packet interacting with single rectangular barrier), subsequently generalizes solve more challenging cases (propagation arbitrarily complex landscape). Furthermore, we demonstrate, using can not only make high-fidelity predictions, also new facts about underlying physical system, detect symmetries, measure relative importance contributing processes.