Quantum Computing: Noise Tailoring Enables 5 Times More Accurate Results with Error Mitigation

The challenge of extracting meaningful results from today’s noisy quantum computers demands innovative error mitigation techniques. Thibault Scoquart from Karlsruhe Institute of Technology, Hugo Perrin from Karlsruhe Institute of Technology and the University of Strasbourg, and Kyrylo Snizhko from Univ. Grenoble Alpes, present a new strategy called Noise Tailoring, which actively reshapes the noise affecting two-qubit gates using statistical sampling. Their research demonstrates that combining Noise Tailoring with existing error mitigation protocols can improve accuracy by up to a factor of five when simulating realistic noise conditions. Importantly, the team also reveals how Noise Tailoring can be repurposed as a diagnostic tool, identifying subtle error sources within quantum hardware to guide future improvements in computer design and performance.

Some error mitigation protocols are particularly efficient for specific types of noise, however the noise in actual hardware may not align with these assumptions. Researchers performed classical emulation of the protocol behaviour and found that the NT+EM results can be up to five times more accurate.

The primary research objective was to investigate whether the statistical properties of noise in quantum circuits could be deliberately altered to improve the performance of existing error mitigation techniques. The study began with classical emulation, where the team engineered a system to statistically sample noise characteristics, effectively reshaping the noise profile prior to applying error mitigation techniques. This approach enables a targeted refinement of noise, optimising it for the specific strengths of chosen EM methods, such as Noise Estimation Circuit (NEC).

Classical simulations demonstrated that combining NT with NEC yielded results up to five times more accurate than utilising NEC alone when simulating realistic Pauli noise acting on two-qubit gates. To validate the technique, scientists transitioned to experiments performed on actual IBM quantum computers, employing the NT+NEC protocol to simulate a quench in the BCS model. The experimental setup involved careful calibration of two-qubit gates and precise control over the statistical sampling process inherent to the NT method. However, results on the quantum hardware revealed a discrepancy, with the NT+NEC protocol performing worse than NEC alone.

Further analysis indicated that the observed performance degradation stemmed from error sources beyond the Markovian Pauli noise accounted for in the classical simulations. The research team harnessed the NT method itself as a diagnostic tool, proposing its use for characterizing these additional error sources present on quantum computers. This innovative application of NT shifts its role from a performance enhancer to a valuable hardware development aid, providing insights into the nuanced noise profiles of NISQ devices. This innovative technique modifies the structure of noise affecting two-qubit gates through a statistical sampling process, aiming to optimise performance when combined with existing error mitigation methods. The research team performed classical emulation to assess the protocol’s behaviour, revealing that integrating NT with the Noise Estimation Circuit (NEC) method can deliver results up to five times more accurate than utilising NEC alone when simulating realistic Pauli noise.

Experiments conducted on actual IBM quantum computers, however, presented a more complex picture. While classical simulations predicted significant gains, the NT+NEC protocol initially yielded poorer results compared to NEC alone. Researchers attribute this discrepancy to the presence of additional, unaccounted-for error sources inherent in the hardware, extending beyond the modelled Markovian Pauli noise. Detailed analysis of the data from these devices revealed that these small error sources are amplified by the NT protocol, impacting overall accuracy. Despite this initial setback, the team discovered that increasing the number of sampling circuits within the NT protocol could improve accuracy by approximately a factor of two compared to the protocol without NT.

Crucially, this in-depth analysis of the NISQ data provided valuable qualitative and quantitative insights into these previously unidentified error sources. The work demonstrates that NT can be repurposed as a powerful diagnostic tool for characterizing subtle errors within quantum hardware. The study proposes leveraging this characterization capability to inform future hardware development, ultimately paving the way for more robust and reliable quantum computers. By employing statistical sampling, NT modifies the structure of noise, and when combined with error mitigation protocols, classical emulation demonstrates significant improvements in accuracy, up to a fivefold increase compared to error mitigation alone, when considering realistic Pauli noise. The method effectively reshapes experimental noise to optimise the performance of existing error mitigation techniques.

The researchers acknowledge that, while promising, the NT method encounters limitations when applied to actual quantum hardware, due to the presence of error sources beyond the initially assumed Markovian Pauli noise. However, they propose leveraging NT as a diagnostic tool for characterizing these additional error sources, potentially informing future hardware development efforts. Future research may focus on extending the protocol to account for more complex noise models and crosstalk effects, and refining the technique to address the discrepancies observed between classical simulations and experimental results.