Effective Noise Thresholds for Reliable Quantum Approximation Optimisation Algorithm under Noisy Environment
Abstract
Variational quantum algorithms are leading candidates for practical utilisation on noisy intermediate-scale quantum (NISQ) processors, but accumulating noise bounds the usefulness of circuit depths. For the quantum approximate optimisation algorithm (QAOA), this trade-off has mostly been addressed by searching for the best depth for a given device. In this work, we consider the inverse question: how much noise each depth can tolerate. Using GPU-accelerated noisy QAOA simulations for MaxCut on three graph families with 6 to 18 vertices, depths up to 10 and three noise channels, we fit the approximation ratios with a generalised logistic surface. We propose the effective noise threshold as the inflection boundary of the logistic surfaces, a noise boundary separating reliable and unreliable QAOA approximation regimes. In general, the threshold decreases as depth rises, indicating that deeper circuits demand for stricter noise requirement for reliable approximation. Its best and worst values over depth divide the noise axis into three regions: every depth remains effective, depth must be chosen with care, or no depth induces effective optimisation. Assessing on the ibm_marrakesh processor, although modelled gate errors understate factual noise, the threshold modeling remains beneficial in informing qualitative depth-dependence structure and identifying dominating noise channel in noisy QAOA.
Related
- Quantum Approximate Optimisation Algorithm for Protein Sidechain Packing
- Endpoint obstructions to upper bounds on Wilsonian thresholds from finite positive effective field theory moments
- Three-Dimensional FDTD Simulation of Superconducting Terahertz Emitters Using an Effective Medium Approximation
- Electrical conductivity of a two-dimensional two-component medium: discretization and effective medium approximation
- A Fully Polynomial Approximation Algorithm for the Sherrington-Kirkpatrick Model
BibTeX
@misc{tjandra2026effective,
title = {{Effective Noise Thresholds for Reliable Quantum Approximation Optimisation Algorithm under Noisy Environment}},
author = {Yozef Tjandra and Hendrik Santoso Sugiarto and Celine Vania Setiadi and Joanna Gracia Tan and Andriyan Bayu Suksmono},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.36597},
eprint = {2610.36597},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.36597},
}
RIS
TY - UNPB
TI - Effective Noise Thresholds for Reliable Quantum Approximation Optimisation Algorithm under Noisy Environment
AU - Yozef Tjandra
AU - Hendrik Santoso Sugiarto
AU - Celine Vania Setiadi
AU - Joanna Gracia Tan
AU - Andriyan Bayu Suksmono
PY - 2026
DA - 2026/10/27
PB - arXiv
DO - 10.48550/arxiv.2610.36597
UR - https://arxiv.org/abs/2610.36597
AB - Variational quantum algorithms are leading candidates for practical utilisation on noisy intermediate-scale quantum (NISQ) processors, but accumulating noise bounds the usefulness of circuit depths. For the quantum approximate optimisation algorithm (QAOA), this trade-off has mostly been addressed by searching for the best depth for a given device. In this work, we consider the inverse question: how much noise each depth can tolerate. Using GPU-accelerated noisy QAOA simulations for MaxCut on three graph families with 6 to 18 vertices, depths up to 10 and three noise channels, we fit the approximation ratios with a generalised logistic surface. We propose the effective noise threshold as the inflection boundary of the logistic surfaces, a noise boundary separating reliable and unreliable QAOA approximation regimes. In general, the threshold decreases as depth rises, indicating that deeper circuits demand for stricter noise requirement for reliable approximation. Its best and worst values over depth divide the noise axis into three regions: every depth remains effective, depth must be chosen with care, or no depth induces effective optimisation. Assessing on the ibm_marrakesh processor, although modelled gate errors understate factual noise, the threshold modeling remains beneficial in informing qualitative depth-dependence structure and identifying dominating noise channel in noisy QAOA.
ER -