UQ-VarQA: Benchmarking and Characterizing NISQ Computers Through Uncertainty Quantification of Variational Quantum Algorithms
Priyabrata Senapati, Shengye Zhu, Bo Peng +2 more
Quantum computing offers speedups, but NISQ processors face hardware-induced errors that degrade fidelity and reproducibility. We introduce an uncertainty-aware benchmarking framework that combines uncertainty quantification with global sensitivity analysis to evaluate not only peak fidelity but also its reliability over time. Using Bayesian Optimization with SGLD refinement under fixed budgets, seeds, bounds, and trust-region rules, and repeating runs across days, we capture calibration drift and quantify efficiency, stability, landscape complexity, and maintenance cost. A noise-aware Gaussian process surrogate provides scalable sensitivity estimates without error mitigation. Applied to VQAMET and VQC on three IBMQ backends, the framework delivers actionable guidance for co-tuning and backend selection, complementing and often outperforming quantum volume style metrics.
IEEE International Conference on Computer Design: VLSI in Computers and Processors (ICCD)•Nov 10, 2025•1 citation