Structure-Aware Minor-Embedding for Machine Learning in Quantum Annealing Processors

Methods for quantum-assisted machine learning (QAML) to train probabilistic models require obtaining high-quality samples from a quantum computing device. In this work we show we can adapt the minor-embedding of Restricted Boltzmann Machines (RBMs) onto qubits in a Quantum Annealing Processor (QAP), to deal with inoperable qubits, and/or to reduce the length of qubit chains. This adaption is possible because connections within an RBM can be omitted without losing overall functionality. These methods for model “pruning” can provide better resource utilization because the reduction of qubit chain length have a direct impact on the quality of samples obtained from the QAP.