Quancorde overview: (1) A canary circuit for the target application is constructed by replacing the non-Clifford gates in the target (orange) with the nearest Clifford gates (green). (2) The target circuit is executed on the diverse noisy quantum ensemble R1-R5 (e.g., different machines or qubits/mappings). (3) In parallel, the canary is also executed on the same noisy ensemble. (4) The correct output of the canary is obtained by running it ideally (noise-free) on a classical machine - possible since Clifford circuits are efficiently classically simulable. (5) Since the correct canary outcome is known, the ensemble is ordered based on the noisy execution fidelity of the canary - R5<R4<R2<R3<R1. (6) The noisy distribution on any machine is selected as the baseline - the correct answer `11' has low probability. (7) Ensemble orderings are produced for the different noisy outputs (of non-negligible probability) of the original circuit and are compared with the canary ordering to estimate a correlation value for each output - `11' has a high correlation. (8) The correlations are then used to weight the baseline distribution to produce a new distribution in which the `11' probability is boosted to become the winner.
On today's noisy imperfect quantum devices, execution fidelity tends to collapse dramatically for most applications beyond a handful of qubits. It is therefore imperative to employ novel techniques that can boost quantum fidelity in new ways. This paper aims to boost quantum fidelity with Clifford canary circuits by proposing Quancorde: Quantum Canary Ordered Diverse Ensembles, a fundamentally newapproach to identifying the correct outcomes of extremely low-fidelity quantum applications. It is based on the key idea of diversity in quantum devices - variations in noise sources, make each (portion of a) device unique, and therefore, their impact on an application's fidelity, also unique. Quancorde utilizes Clifford canary circuits (which are classically simulable, but also resemble the target application structure and thus suffer similar structural noise impact) to order a diverse ensemble of devices or qubits/mappings approximately along the direction of increasing fidelity of the target application. Quancorde then estimates the correlation of the ensemble-wide probabilities of each output string of the application, with the canary ensemble ordering, and uses this correlation to weight the application's noisy probability distribution. The correct application outcomes are expected to have higher correlation with the canary ensemble order, and thus their probabilities are boosted in this process. Doing so, Quancorde improves the fidelity of evaluated quan-tum applications by a mean of 8.9x/4.2x (wrt. different baselines) and up to a maximum of 34x. DOI 10.1109/ICRC57508.2022.00014
G. S. Ravi, J. M. Baker, K. N. Smith, N. Earnest, A. Javadi-Abhari and F. T. Chong
