Machine learning, quantum chaos, and pseudorandom evolution

Abstract

By modeling quantum chaotic dynamics with ensembles of random operators, we explore howmachine learning learning algorithms can be used to detect pseudorandom behavior in qubit systems.We analyze samples consisting of pieces of correlation functions and find that machine learningalgorithms are capable of determining the degree of pseudorandomness which a system is subjectto in a precise sense. This is done without computing any correlators explicitly. Interestingly,even samples drawn from two-point functions are found to be sufficient to solve this classificationproblem. This presents the possibility of using deep learning algorithms to explore late time behaviorin chaotic quantum systems which have been inaccessible to simulation.