ľ�ϸ���Ӱ��

Title: Automatic differentiation: a one-fits-all solution in diffractive imaging

Abstract: In this talk, I will provide a broad overview of the world of Automatic Differentiation Ptychography (ADP), a lensless imaging method that allows for wavefront sensing and phase-sensitive microscopy. Ptychography is widely used to characterize structured samples from measurements of diffraction intensity patterns. To solve the optimization task in the inverse problem posed by this method, we employ automatic differentiation (AD) to build a flexible and extendable framework. I will highlight the capabilities of our approach to retrieve optical design parameters, precise multi-parameter estimations, and minimize a Fisher information metric using deep learning libraries—all of this with relatively low implementation labor.
I will then consider a real-space parameterization of an object composed of three lines and demonstrate how to identify the optimal illumination scheme that minimizes the estimation error under specified experimental constraints. These results offer new insights to improve the performance of methods based on sparsity-based coherent diffraction imaging when the dose per acquisition may be limited, notably for the characterization of delicate samples or when high throughput is required.

Article on which your lecture is based: Bouchet, D., Seifert, J. & Mosk, A. P. “Optimizing illumination for precise multi-parameter estimations in coherent diffractive imaging”. Opt. Lett., OL 46, 254–257 (2021), DOI: