Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Classical Noise Inversion: A Practical and Optimal framework for Robust Quantum Applications

Dayue Qin, Ying Li, You Zhou

TL;DR

The paper introduces Classical Noise Inversion (CNI), a framework that shifts quantum error mitigation from quantum circuits to classical post-processing by inverting propagated noise using a condition that enables measurement channels to be bypassed through classical computation. By pairing CNI with Noise Compression and integrating with shadow estimation, the authors achieve unbiased estimates under gate-dependent noise with substantially reduced variance and sampling overhead compared to conventional probabilistic error cancellation. Theoretical results establish the validity via Pauli twirling and a compressed inverse that minimizes classical overhead, while practical implementations (Local and Global) provide scalable routes for both uncorrelated and correlated noise. Numerical simulations demonstrate robust performance on small-scale quantum tasks, highlighting a path toward scalable, practical quantum applications on noisy devices and offering a versatile framework for a broad class of classically simulable quantum protocols.

Abstract

Quantum error mitigation is a critical technology for extracting reliable computations from noisy quantum processors, proving itself essential not only in the near term but also as a valuable supplement to fully fault-tolerant systems in the future. However, its practical implementation is hampered by two major challenges: the expansive cost of sampling from quantum circuits and the reliance on unrealistic assumptions, such as gate-independent noise. Here, we introduce Classical Noise Inversion (CNI), a framework that fundamentally bypasses these crucial limitations and is well-suited for various quantum applications. CNI effectively inverts the accumulated noise entirely during classical post-processing, thereby eliminating the need for costly quantum circuit sampling and remaining effective under the realistic condition of gate-dependent noise. Apart from CNI, we introduce noise compression, which groups noise components with equivalent effects on measurement outcomes, achieving the optimal overhead for error mitigation. We integrate CNI with the framework of shadow estimation to create a robust protocol for learning quantum properties under general noise. Our analysis and numerical simulations demonstrate that this approach substantially reduces statistical variance while providing unbiased estimates in practical situations where previous methods fail. By transforming a key quantum overhead into a manageable classical cost, CNI opens a promising pathway towards scalable and practical quantum applications.

Classical Noise Inversion: A Practical and Optimal framework for Robust Quantum Applications

TL;DR

The paper introduces Classical Noise Inversion (CNI), a framework that shifts quantum error mitigation from quantum circuits to classical post-processing by inverting propagated noise using a condition that enables measurement channels to be bypassed through classical computation. By pairing CNI with Noise Compression and integrating with shadow estimation, the authors achieve unbiased estimates under gate-dependent noise with substantially reduced variance and sampling overhead compared to conventional probabilistic error cancellation. Theoretical results establish the validity via Pauli twirling and a compressed inverse that minimizes classical overhead, while practical implementations (Local and Global) provide scalable routes for both uncorrelated and correlated noise. Numerical simulations demonstrate robust performance on small-scale quantum tasks, highlighting a path toward scalable, practical quantum applications on noisy devices and offering a versatile framework for a broad class of classically simulable quantum protocols.

Abstract

Quantum error mitigation is a critical technology for extracting reliable computations from noisy quantum processors, proving itself essential not only in the near term but also as a valuable supplement to fully fault-tolerant systems in the future. However, its practical implementation is hampered by two major challenges: the expansive cost of sampling from quantum circuits and the reliance on unrealistic assumptions, such as gate-independent noise. Here, we introduce Classical Noise Inversion (CNI), a framework that fundamentally bypasses these crucial limitations and is well-suited for various quantum applications. CNI effectively inverts the accumulated noise entirely during classical post-processing, thereby eliminating the need for costly quantum circuit sampling and remaining effective under the realistic condition of gate-dependent noise. Apart from CNI, we introduce noise compression, which groups noise components with equivalent effects on measurement outcomes, achieving the optimal overhead for error mitigation. We integrate CNI with the framework of shadow estimation to create a robust protocol for learning quantum properties under general noise. Our analysis and numerical simulations demonstrate that this approach substantially reduces statistical variance while providing unbiased estimates in practical situations where previous methods fail. By transforming a key quantum overhead into a manageable classical cost, CNI opens a promising pathway towards scalable and practical quantum applications.
Paper Structure (25 sections, 6 theorems, 96 equations, 6 figures, 1 algorithm)

This paper contains 25 sections, 6 theorems, 96 equations, 6 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Results
  3. Theoretical framework of classical noise inversion
  4. Validity and optimality of CNI
  5. CNI-based robust quantum learning
  6. Numerical simulations
  7. Discussion
  8. Methods
  9. Practical implementation
  10. Local approach
  11. Global approach
  12. Efficient $\mathbb{Z}$-block comparison
  13. Performance of classical noise inversion
  14. Performance of CNI
  15. Performance of conventional error mitigation
  16. ...and 10 more sections

Key Result

Theorem 1

With one-shot noisy measurement outcome $b$ and one instance of a random basis channel $\mathcal{B}$, The unbiased estimator of the expectation value $f$ in Eq. eq:f is given by Suppose that the noisy circuit is measured for $K$ shots. For each shot, $L$ instances of $\mathcal{B}$ are generated, yielding $K\times L$ estimators of $f$ constructed through eq:hatf in total. Let $\hat{f}_{\mathrm{tot

Figures (6)

  • Figure 1: Illustration of the framework of CNI and its applications. (a) The scheme CNI incorporated with noise compression. Both CNI and noise compression are implemented on a classical computer. The compression procedure transforms the actual hard-to-mitigate noise into easy-to-mitigate noise, further reducing the cost of QEM. Then, the noise effect is canceled by classically simulating the noise inversion using efficient methods such as Monte-Carlo or tensor-network-based methods. (b) Representative quantum applications amenable to the CNI approach. (c) Flowchart of the CNI-based robust shadow estimation protocol.
  • Figure 2: A two qubit example of twirling and compression of noise models. (a) The Pauli transfer matrix of general quantum noise operation. Where every element of the matrix (plaquettes filled with color) can be non-zero (since we impose no physical constraints on the noise). (b) The $\mathbb{Z}$-twirled Pauli transfer matrix. Some of the elements of the original matrix are transformed to zero (empty plaquettes). According to Proposition \ref{['prop:valid']}, the upper right block being zero is the necessary and sufficient condition for CNI validity. (c) After both $\mathbb{Z}$- and $\mathbb{X}$-twirling, i.e. Pauli twirling, the Pauli transfer matrix is diagonal. (d, e) Noise compression (Comp.) from the perspective in the Pauli transfer matrix, we only need to consider the $\mathbb{Z}$-block, and neglect the non-$\mathbb{Z}$-block elements (plaquettes filled with gray).
  • Figure 3: Numerical comparison of our method and existing methods under multi-shot measurement scheme. In the figure, CNI stands for the proposed CNI-based robust shadow estimation, sRSE stands for standard robust shadow estimation and cPEC stands for conventional probabilistic error cancellation. The fidelity of $4$-qubit GHZ state is estimated at a cost of $M=10^3,K=10^3,L=1$ across varying error rates; this estimation procedure is repeated for 320 times. Data points represent the mean values of the repeated estimations, while the shaded areas indicates one standard deviation. The standard deviations are also plotted separately in the inset. The calibration cost for sRSE is $M=3.2\times 10^5, K=1,L=1$. (a) Local approach for uncorrelated noise. (b) Global approach for correlated noise.
  • Figure 4: Numerical comparison of our method and existing method under single-shot measurement scheme. The details of the plots are identical to those in Fig. \ref{['fig: numerical main-MS']} except that the estimation cost is set to $M=10^3,K=1,L=1$. Additionally, the results of CNI with an estimation cost of $M=10^3,K=1,L=10$ are included.
  • Figure 5: Illustration of practical implementation of twirling. (a) Local twirling for uncorrelated noise. (b) global twirling for correlated noise.
  • ...and 1 more figures

Theorems & Definitions (7)

  • Theorem 1: Performance guarantee of CNI
  • Proposition 1
  • Proposition 2
  • Theorem 2: Performance guarantee of CNI-based robust shadow estimation
  • Proposition 3: Upper bound of noisy shadow norm
  • Theorem 3
  • proof