Quantum neural compressive sensing for ghost imaging

TL;DR

This paper introduces a quantum neural compressive sensing algorithm for ghost imaging that leverages variational quantum circuits and physical models to outperform traditional methods, demonstrating robustness and potential for near-term quantum devices.

Contribution

It presents a novel quantum neural network approach for ghost imaging that incorporates physical inductive bias to enhance performance and overcome optimization challenges.

Findings

Outperforms conventional ghost imaging methods in visual and quantitative metrics

Effectively overcomes barren plateau problem in quantum optimization

Robust against various quantum noise levels

Abstract

Demonstrating the utility of quantum algorithms is a long-standing challenge, where quantum machine learning becomes one of the most promising candidate that can be resorted to. In this study, we investigate a quantum neural compressive sensing algorithm for ghost imaging to showcase its utility. The algorithm utilizes the variational quantum circuits to reparameterize the inverse problem of ghost imaging and uses the inductive bias of the physical forward model to perform optimization. To validate the algorithm's effectiveness, we conduct optical ghost imaging experiments, capturing signals from objects at different physical sampling rates and detection signal-to-noise ratios. The experimental results show that our proposed algorithm surpasses conventional methods in both visual appearance and quantitative metrics, achieving state-of-the-art performance. Importantly, we observe that the quantum neural network, guided by prior knowledge of physics, effectively overcomes the challenge of barren plateau in the optimization process. The proposed algorithm demonstrates robustness against various quantum noise levels, making it suitable for near-term quantum devices. Our study leverages physical inductive bias guided variational quantum algorithm, underscoring the potential of quantum computation in tackling a broad range of optimization and inverse problems.