Improving success probability in the LHZ parity embedding by computing with quantum walks

TL;DR

This paper investigates the effectiveness of continuous-time quantum walks in the LHZ parity embedding for quantum optimization, demonstrating improved success probabilities through error correction and validating theoretical bounds.

Contribution

It explores the combination of quantum walks with LHZ embedding, providing numerical evidence and identifying error correction methods to enhance success rates.

Findings

Quantum walks perform well with LHZ embedding on small instances.

Optimal hopping rates can be estimated using existing heuristics.

Error correction techniques improve success probability.

Abstract

The LHZ parity embedding is one of the front-running methods for implementing difficult-to-engineer long-range interactions in quantum optimisation problems. Continuous-time quantum walks are a leading approach for solving quantum optimisation problems. Due to them populating excited states, quantum walks can avoid the exponential gap closing problems seen in other continuous-time techniques such as quantum annealing and adiabatic quantum computation (AQC). An important question therefore, is how continuous-time quantum walks perform in combination with the LHZ parity embedding. By numerically simulating continuous-time quantum walks on 4, 5 and 6 logical qubit Sherrington-Kirkpatrick (SK) Ising spin glass instances embedded onto the LHZ parity architecture, we are able to verify the continued efficacy of heuristics used to estimate the optimal hopping rate and the numerical agreement with the theory behind the location of the lower bound of the LHZ parity constraint strength. In addition, by comparing several different LHZ-based decoding methods, we were able to identify post-readout error correction techniques which were able to improve the success probability of the quantum walk.