Machine learning the dynamics of quantum kicked rotor

Using the multilayer convolutional neural network (CNN), we can detect quantum phases in random electron systems, and phase diagrams of two higher dimensional Anderson transitions percolations as well disordered topological systems have been obtained. Here, instead using CNN to analyze wave functions, dynamics packets via long short-term memory (LSTM). We adopt quasi-periodic kicked rotors, which simulate three four transitions. By supervised training, let LSTM extract features time series packet displacements localized delocalized phases. then unknown classify compare obtained by those CNN.