Error mitigation by training with fermionic linear optics
AI Breakdown
Get a structured breakdown of this paper — what it's about, the core idea, and key takeaways for the field.
Abstract
Noisy intermediate-scale quantum (NISQ) computers could solve quantum-mechanical simulation problems that are beyond the capabilities of classical computers. However, NISQ devices experience significant errors which, if not corrected, can render physical quantities measured in these simulations inaccurate or meaningless. Here we describe a method of reducing these errors which is tailored to quantum algorithms for simulating fermionic systems. The method is based on executing quantum circuits in the model of fermionic linear optics, which are known to be efficiently simulable classically, to infer the relationship between exact and noisy measurement outcomes, and hence undo the effect of noise. We validated our method by applying it to the VQE algorithm for estimating ground state energies of instances of the Fermi-Hubbard model. In classical numerical simulations of 12-qubit examples with physically realistic levels of depolarising noise, errors were reduced by a factor of around 34 compared with the uncorrected case. Smaller experiments on quantum hardware demonstrate an average reduction in errors by a factor of 10 or more.