Neural Error Mitigation of Near-Term Quantum Simulations

Near-term quantum computers provide a promising platform for finding the ground states of systems, which is an essential task in physics, chemistry and materials science. However, near-term approaches are constrained by effects noise, as well limited resources hardware. We introduce neural error mitigation, uses networks to improve estimates ground-state observables obtained using simulations. To demonstrate our method’s broad applicability, we employ mitigation find H2 LiH molecular Hamiltonians, lattice Schwinger model, prepared via variational eigensolver. Our results show that improves numerical experimental eigensolver computations yield low energy errors, high fidelities accurate estimations more complex such order parameters entanglement entropy without requiring additional resources. Furthermore, agnostic with respect state preparation algorithm used, hardware it implemented on particular noise channel affecting experiment, contributing its versatility tool simulation. So-called noisy intermediate-scale devices will be capable range simulation tasks, provided can sufficiently reduced. A approach developed molecules systems simulations devices.