Southwest Quantum Information and Technology

Variationally scheduled adiabatic quantum computing

Presenting Author: Shunji Matsuura, 1QBit

Solving quantum and classical optimization problems are one of the major applications of quantum computing. Some important limitations NISQ devices are the absence of error correction. Therefore, shortening the time taken to perform a single run of a quantum algorithm is essential for successfully obtaining accurate results. In this work, we consider AQC approach. In AQC, computational results highly depend on the annealing time and the annealing path. While a long annealing time causes accumulation of environment-induced errors, a small annealing time may cause harmful diabatic transitions. In order to avoid those problems, we propose to find optimal annealing paths variationally. We consider two variational methods. The first method introduces a new Hamiltonian which is characterized by variational parameters and has a finite coefficient only in the middle of annealing. The second method uses independent schedule functions for terms in the Hamiltonian. These schedule functions are determined variationally. In both methods, the required annealing time per single run can be reduced by some orders of magnitude. The major findings are: 1) Our variational methods provide accurate results even in the presence of noise, while the standard AQC fails to do so. 2) Our variational methods provide accurate results even when the final Hamiltonian on a quantum device has inaccurate couplings. 3) It can reduce the total annealing time compared to the standard AQC.

(Session 5 : Saturday from 5:00pm - 7:00pm)