【摘 要】
：
The quality of inverse problem solutions obtained through deep learning is limited by the nature of the priors learned from examples presented during the traini
【机 构】
：
Department of Electrical Engineering and Computer Science,Massachusetts Institute of Technology,Camb
论文部分内容阅读
The quality of inverse problem solutions obtained through deep learning is limited by the nature of the priors learned from examples presented during the training phase. Particularly in the case of quantitative phase retrieval, spatial frequencies that are underrepresented in the training database, most often at the high band, tend to be suppressed in the reconstruction. Ad hoc solutions have been proposed, such as pre-amplifying the high spatial frequencies in the examples;however, while that strategy improves the resolution, it also leads to high-frequency artefacts, as well as low-frequency distortions in the reconstructions. Here, we present a new approach that learns separately how to handle the two frequency bands, low and high, and learns how to synthesize these two bands into full-band reconstructions. We show that this“learning to synthesize”(LS) method yields phase reconstructions of high spatial resolution and without artefacts and that it is resilient to high-noise conditions, e.g., in the case of very low photon flux. In addition to the problem of quantitative phase retrieval, the LS method is applicable, in principle, to any inverse problem where the forward operator treats different frequency bands unevenly, i.e., is ill-posed. This work was supported by the Intelligence Advanced Research Projects Activity (IARPA) grant No. FA8650-17-C-9113 and by the SenseTime company. I.K. was supported in part by the KFAS (Korea Foundation for Advanced Studies) scholarship. We are grateful to Kwabena Arthur and Maciej Baranski for useful discussions and critiques of the paper.
其他文献
A four-fold-degenerate three-dimensional (3D) Dirac point, represents a degenerate pair of Weyl points carrying opposite chiralities. Moreover, 3D Dirac crystal
The photothermoelectric (PTE) effect enables efficient harvesting of the energy of photogenerated hot carriers and is a promising choice for high-efficiency pho
The multi-dimensional laser is a fascinating platform not only for the discovery and understanding of new higher-dimensional coherent lightwaves but also for th
The success of photonic crystal fibres relies largely on the endless variety of two-dimensional photonic crystals in the cross-section. Here, we propose a topol
Long-range and fast transport of coherent excitons is important for the development of high-speed excitonic circuits and quantum computing applications. However
The flourishing of topological photonics in the last decade was achieved mainly due to developments in linear topological photonic structures. However, when non
The launch of the big data era puts forward challenges for information preservation technology, both in storage capacity and security. Herein, a brand new optic
Chemical imaging based on mid-infrared (MIR) spectroscopic contrast is an important technique with a myriad of applications, including biomedical imaging and en
Efficient calculation of the light diffraction in free space is of great significance for tracing electromagnetic field propagation and predicting the performan
Lateral optical forces induced by linearly polarized laser beams have been predicted to deflect dipolar particles with opposite chiralities toward opposite tran