Quantum Natural Gradient optimizer on noisy platforms: QAOA as a case study

TL;DR

This paper evaluates the Quantum Natural Gradient optimizer's performance in noisy quantum environments, demonstrating its superior convergence and robustness over traditional methods within the QAOA framework on real quantum hardware.

Contribution

It provides the first comprehensive benchmarking of QNG against vanilla gradient descent in noisy quantum platforms for QAOA, highlighting its advantages in robustness and convergence.

Findings

QNG outperforms vanilla gradient descent in convergence speed.

QNG shows greater robustness against noise and initial parameter variations.

QNG effectively avoids local minima, improving optimization in NISQ devices.

Abstract

We investigate the performance of the Quantum Natural Gradient (QNG) optimizer in the presence of noise. Specifically, we evaluate the efficacy of QNG within the Quantum Approximate Optimization Algorithm (QAOA) for finding the ground state of the Transverse Field Ising Model (TFIM). Its performance is benchmarked against the Vanilla Gradient Descent optimizer across two prominent quantum computing platforms: Rydberg atoms and superconducting circuits. Our analysis includes simulations under both idealized noise-free conditions and realistic noisy environments based on calibration data from actual devices. Results demonstrate that QNG consistently outperforms Vanilla Gradient Descent, exhibiting faster convergence on average and greater robustness against random initializations of parameters. This robustness is attributed to the distance regularization in parameter space inherent to QNG. Additionally, QNG achieves a higher convergence rate to the solution, effectively avoiding certain local minima. These findings highlight QNG as a promising tool for optimizing variational quantum algorithms in noisy intermediate-scale quantum (NISQ) devices.