Make Some Noise! Measuring Noise Model Quality in Real-World Quantum Software

TL;DR

This paper develops and validates a tunable noise model for quantum simulators based on empirical data from IBM hardware, enabling realistic simulation and assessment of noise model quality for current and future quantum algorithms.

Contribution

It introduces a scalable, empirically calibrated noise model using Kraus channels, validated on large-scale simulators and real hardware, with a new metric for noise model quality assessment.

Findings

Accurately simulates IBM quantum hardware noise

Provides a method to evaluate noise model quality for large problems

Offers insights into noise effects on future fault-tolerant quantum systems

Abstract

Noise and imperfections are among the prevalent challenges in quantum software engineering for current NISQ systems. They will remain important in the post-NISQ area, as logical, error-corrected qubits will be based on software mechanisms. As real quantum hardware is still limited in size and accessibility, noise models for classical simulation--that in some cases can exceed dimensions of actual systems--play a critical role in obtaining insights into quantum algorithm performance, and the properties of mechanisms for error correction and mitigation. We present, implement and validate a tunable noise model building on the Kraus channel formalism on a large scale quantum simulator system (Qaptiva). We use empirical noise measurements from IBM quantum (IBMQ) systems to calibrate the model and create a realistic simulation environment. Experimental evaluation of our approach with Greenberger-Horne-Zeilinger (GHZ) state preparation and QAOA applied to an industrial use-case validate our approach, and demonstrate accurate simulation of hardware behaviour at reasonable computational cost. We devise and utilise a method that allows for determining the quality of noise models for larger problem instances than is possible with existing metrics in the literature. To identify potentials of future quantum software and algorithms, we extrapolate the noise model to future partially fault-tolerant systems, and give insights into the interplay between hardware-specific noise modelling and hardware-aware algorithm development.