Quantum-assisted Helmholtz machines: A quantum–classical deep learning framework for industrial datasets in near-term devices

Abstract

Machine learning has been presented as one of the key applications for\nnear-term quantum technologies, given its high commercial value and wide range\nof applicability. In this work, we introduce the \\textit{quantum-assisted\nHelmholtz machine:} a hybrid quantum-classical framework with the potential of\ntackling high-dimensional real-world machine learning datasets on continuous\nvariables. Instead of using quantum computers only to assist deep learning, as\nprevious approaches have suggested, we use deep learning to extract a\nlow-dimensional binary representation of data, suitable for processing on\nrelatively small quantum computers. Then, the quantum hardware and deep\nlearning architecture work together to train an unsupervised generative model.\nWe demonstrate this concept using 1644 quantum bits of a D-Wave 2000Q quantum\ndevice to model a sub-sampled version of the MNIST handwritten digit dataset\nwith 16x16 continuous valued pixels. Although we illustrate this concept on a\nquantum annealer, adaptations to other quantum platforms, such as ion-trap\ntechnologies or superconducting gate-model architectures, could be explored\nwithin this flexible framework.\n