NBIA/QDev Seminar: Troels Frimodt Rønnow

Beyond hundred qubits: It's just classical, right?

Abstract
We perform experiments on the 128 qubit annealing processor, D- Wave One, comparing thousands of instances with simulated classical and quantum annealing. Within the processors noise and calibration uncertainties, we nd that the results generated by the quantum annealer are statistically indistinguishable from results generated by a simulated quantum annealer while signicantly dierent from those of a classical annealer. This demonstrates that the device performs quantum annealing. An intriguing feature is strong bimodal separation of the instances of intermediate sizes into two categories: hard and easy, which is not observed for the classical annealer.

Further, we investigate hardness and scaling of \time-to-solution" for several thousand realisations of J spin glass problems, ranging from 8 to 108 qubits in size. We demonstrate that using gauge invariant transformations is a crucial optimisation technique to compensate for calibration uncertainties. Based on the similarities between the simulated quantum annealer and D-Wave One, we make predictions for the 512 qubit processor, D-Wave Two, using our simulated quantum annealers and provide reference data against which quantum speedup could be observed.