Benchmarking Amplitude Estimation on a Superconducting Quantum Computer

TL;DR

This paper benchmarks amplitude estimation on a superconducting quantum computer, demonstrating that maximum likelihood estimation with optimized circuits outperforms naive sampling up to a certain circuit depth, informing progress toward quantum advantage.

Contribution

It provides empirical upper bounds on feasible circuit depths for amplitude estimation on current superconducting quantum hardware using MLE.

Findings

MLE with optimized circuits outperforms naive sampling up to 3 Grover iterations

Circuit depth of 131 is achievable with current hardware for amplitude estimation

Benchmarking progress towards quantum advantage on NISQ devices

Abstract

Amplitude Estimation (AE) is a critical subroutine in many quantum algorithms, allowing for a quadratic speedup in various applications like those involving estimating statistics of various functions as in financial Monte Carlo simulations. Much work has gone into devising methods to efficiently estimate the amplitude of a quantum state without expensive operations like the Quantum Fourier Transform (QFT), which is especially prohibitive given the constraints of current NISQ devices. Newer methods have reduced the number of operations required on a quantum computer and are the most promising near-term implementations of the AE subroutine. While it remains to be seen the exact circuit requirements for a quantum advantage in applications relying on AE, it is necessary to continue to benchmark the algorithm's performance on current quantum computers and the circuit costs associated with such subroutines. Given these considerations, we expand on results from previous experiments in using Maximum Likelihood Estimation (MLE) to approximate the amplitude of a quantum state and provide empirical upper bounds on the current feasible circuit depths for AE on a superconducting quantum computer. Our results show that MLE using optimally-compiled circuits can currently outperform naive sampling for up to 3 Grover Iterations with a circuit depth of 131, which is higher than reported in other experimental results. This functional benchmark is one of many that will be continually monitored against current quantum hardware to measure the necessary progress towards quantum advantage.