Machine Learning Regression of Quantum Many-Body Operator Dynamics

The accurate determination of the long-time dynamics of operator expectation values for quantum many body systems is a computationally demanding problem, with traditional methods scaling exponentially with the system size. We develop a machine learning method which determines the long time dynamics by performing a regression over expectation values calculated exactly over short time intervals. WIth this approach, the long-time dynamics can be determined independent of system size. We demonstrate this computational advantage for both the Ising model in transverse field and the XXZ model.