Hybrid quantum-classical generative adversarial networks for image generation via learning discrete distribution

It has been reported that quantum generative adversarial networks have a potential exponential advantage over classical generative adversarial networks. However, quantum machine learning is difficult to find real applications in the near future due to the limitation of quantum devices. The structure of quantum generator is optimized to reduce the required parameters and make use of quantum devices to a greater extent. And an image generation scheme is designed based on quantum generative adversarial networks. Two structures of quantum generative adversarial networks are simulated on Bars and Stripes dataset, and the results corroborate that the quantum generator with reduced parameters has no visible performance loss. The original complex multimodal distribution of an image can be converted into a simple unimodal distribution by the remapping method. The MNIST images and the Fashion-MNIST images are successfully generated by the optimized quantum generator with the remapping method, which verified the feasibility of the proposed image generation scheme.

Automated summary

It has been reported that Quantum Generative Adversarial Networks (QGANs) have a potential exponential advantage over classical Generative Adversarial Networks (GANs). However, the limitation of current quantum devices makes it difficult for quantum machine learning to find practical applications in the near future. To address this, the structure of the quantum generator is optimized to reduce parameters and leverage quantum devices effectively. An image generation scheme based on QGANs is proposed and validated on Bars and Stripes dataset, showing that the optimized quantum generator maintains performance without visible losses. Furthermore, the feasibility of the proposed scheme is demonstrated by successfully generating MNIST and Fashion-MNIST images using the optimized quantum generator and remapping method.