Machine Learning Augmented Shadow Tomography (Part II)

Building on the discussion of the two elements of the Machine-learning Augmented Shadow Tomography (MAST) presented in the first talk, we present results of using MAST on estimation tasks relying on experimentally accessible measurements. Specifically, we consider estimation performance on GHZ, Haar random, and Bell pair product states. We benchmark the performance of MAST against classical shadow without data augmentation. We discuss how these results motivate the application of MAST to experimental systems with sparse measurements.