Optimized Circuit Cutting for QAOA Sampling Tasks

TL;DR

This paper explores applying circuit cutting to QAOA sampling tasks, demonstrating that optimized cutting schemes can scale problem size and reduce noise effects, despite some distribution shifts towards suboptimal solutions.

Contribution

It introduces a method to minimize cuts via integer programming to balance distribution accuracy and noise reduction in quantum sampling.

Findings

Circuit cutting broadens and shifts the bitstring distribution towards suboptimal values.

Optimized cutting schemes reduce circuit size and noise impact on hardware.

Noise reduction benefits outweigh distribution degradation for large circuits.

Abstract

Circuit cutting was originally designed to retrieve the expectation value of an observable with respect to a large quantum circuit by executing smaller circuit fragments. In this work, however, we demonstrate the application of circuit cutting to a pure sampling task. In particular, we sample solutions to an optimization problem from a trained QAOA circuit. Here, circuit cutting leads to a broadening and shift of the bitstring distribution towards suboptimal values compared to the uncut case. To reduce this effect, we minimize the number of required cuts via integer programming methods. On the other hand, cutting reduces the circuit size and thus the impact of noise. Our experiments on quantum hardware reveal that, for large circuits, the effect of noise reduction outweighs the derogative effects on the bitstring distribution. The study therefore provides evidence that circuit cutting combined with optimized cutting schemes can both scale problem size and mitigate noise for near-term quantum optimization.