Abstract

We integrate quantum computing with GANs into a hybrid quantum-classical GAN to generate complex 5-channel electron backscatter diffraction (EBSD) images of two distinct microstructure phases of steel. By training a quantum circuit at the input layer of a large classical Wasserstein GAN (WGAN) model, we achieve higher image quality compared to a baseline classical GAN. We generate images from ferrite and bainite microstructure phases and evaluate performance with maximum mean discrepancy, finding that the hybrid model improves over classical Bernoulli GANs in 70% of samples. The method is demonstrated on both quantum simulators and trapped-ion quantum hardware.

Abstract

We integrate quantum computing with GANs into a hybrid quantum-classical GAN to generate complex 5-channel electron backscatter diffraction (EBSD) images of two distinct microstructure phases of steel. By training a quantum circuit at the input layer of a large classical Wasserstein GAN (WGAN) model, we achieve higher image quality compared to a baseline classical GAN. We generate images from ferrite and bainite microstructure phases and evaluate performance with maximum mean discrepancy, finding that the hybrid model improves over classical Bernoulli GANs in 70% of samples. The method is demonstrated on both quantum simulators and trapped-ion quantum hardware.

Abstract

We integrate quantum computing with GANs into a hybrid quantum-classical GAN to generate complex 5-channel electron backscatter diffraction (EBSD) images of two distinct microstructure phases of steel. By training a quantum circuit at the input layer of a large classical Wasserstein GAN (WGAN) model, we achieve higher image quality compared to a baseline classical GAN. We generate images from ferrite and bainite microstructure phases and evaluate performance with maximum mean discrepancy, finding that the hybrid model improves over classical Bernoulli GANs in 70% of samples. The method is demonstrated on both quantum simulators and trapped-ion quantum hardware.

Abstract

We integrate quantum computing with GANs into a hybrid quantum-classical GAN to generate complex 5-channel electron backscatter diffraction (EBSD) images of two distinct microstructure phases of steel. By training a quantum circuit at the input layer of a large classical Wasserstein GAN (WGAN) model, we achieve higher image quality compared to a baseline classical GAN. We generate images from ferrite and bainite microstructure phases and evaluate performance with maximum mean discrepancy, finding that the hybrid model improves over classical Bernoulli GANs in 70% of samples. The method is demonstrated on both quantum simulators and trapped-ion quantum hardware.