Benchmarking of quantum generative adversarial networks using NVIDIA’s Quantum Optimized Device Architecture

Quantum generative adversarial models (QGANs) have the potential to vastly improve the training of machine learning models by providing accelerated learning and stronger expressivity compared to classical GANs. In this study, we present our results from benchmarking a GPU accelerated hybrid QGAN with a quantum generator and a classical discriminator using Nvidia’s Quantum-Optimized Device Architecture (QODA). QODA provides a heterogeneous quantum-classical workflow that is ideal for such applications. Its modern C++ based programming model is designed for interoperability with existing classical parallel programming models.