Natural evolutionary strategies for variational quantum computation

Abstract Natural evolutionary strategies (NES) are a family of gradient-free black-box optimization algorithms. This study illustrates their use for the randomly initialized parameterized quantum circuits (PQCs) in region vanishing gradients. We show that using NES gradient estimator exponential decrease variance can be alleviated. implement two specific approaches, and separable NES, parameter PQCs compare them against standard descent. apply to different problems ground state energy estimation variational eigensolver preparation with varying depth length. also introduce batch larger extend ES number parameters. achieve accuracy comparable state-of-the-art techniques all above cases lower circuit evaluations. Our empirical results indicate one as hybrid tool tandem other gradient-based methods deep regions