2023/10/16 by Gino Bishop, Bishop, Gino, Simone Montangero +3
Computer Science · Decision Sciences · Materials Science · #FOS: Physical sciences #Machine Learning in Materials Science #Neural Networks and Reservoir Computing #Quantum Physics (quant-ph) #Scientific Computing and Data Management
paper · pdf · doi:10.48550/arxiv.2310.10442
openalex publication_date 2023/10/16 · openalex created_date 2023/10/18 · openalex updated_date 2026/07/28
We perform exact diagonalization and time evolution of the Lechner-Hauke-Zoller (LHZ) annealing architecture [Science Advances 1(9), e1500838 (2015)] for ten physical qubits. Thereby, on a training set consisting of 2400 problem instances, we perform the optimization task of tuning the local fields with the goal to identify a set of fixed optimal annealing protocols, that outperforms linear protocols on a large class of arbitrary LHZ problem instances. We show that average ground state fidelities of 0.9 can be achieved by applying optimized protocols onto groups of problem instances with similar energy landscapes. Particularly, the set of optimized annealing protocols reduces annealing time required to reach a predefined threshold ground state fidelity by an average of ∼72%, corresponding to a speed-up of factor ∼3.5. Moreover, as these protocols are meant to be readily applicable in experimental setups, they can be used as a starting point for problem-specific protocol optimization. In reverse, we discuss how previously optimized protocols can potentially be used to gauge the instantaneous energy landscape of a spin glass problem. Albeit simulations were performed on the LHZ architecture, identification of optimized protocols is not limited to either simulations or local connectivity only.