Faster Probabilistic Error Cancellation

TL;DR

This paper introduces a new probabilistic error cancellation method that reduces sampling costs by decomposing inverse channels and reorganizing circuits, leading to more efficient quantum error mitigation with controlled bias.

Contribution

The work presents a novel PEC approach that reorganizes circuits based on inverse generator powers, achieving lower sampling overhead and better bias control than standard PEC methods.

Findings

Analytical and numerical evidence of reduced sampling cost.

Experimental validation showing agreement with ideal values.

Outperforms heuristic methods like zero-noise extrapolation in certain scenarios.

Abstract

Probabilistic error cancellation (PEC) is a leading quantum error mitigation method that provides an unbiased estimate, although it is known to have a large sampling overhead. In this work, we propose a new method to perform PEC, which results in a lower sampling cost than the standard way. It works by decomposing the inverse channel of each gate or each circuit layer into the identity part and the non-identity part and reorganizing the full circuit as different powers of the inverse generator. The ideal circuit becomes a linear combination of noisy circuits with different weights where shots are deterministically allocated to each circuit based on its weight. This naturally sets the achievable bias given a finite amount of shots. As the number of shots is increased, smaller bias terms can be gradually resolved and become bias-free in the limit of sufficient shots. We show the saving both analytically and numerically over the standard PEC and identify situations where it can outperform heuristic approach, such as zero-noise extrapolation, due to the well-controlled bias. We also demonstrated this method experimentally and found excellent agreement between the mitigated and the ideal values.