报告题目：Mixing and demixing optical information: Reconstructive Spectrometer & Deep Learning-enabled Holography

报告人：Mooseok Jang 副教授

报告人单位：Korea Advanced Institute of Science and Technology

主持人：沈乐成 教授

报告时间：2026年10月26日（周一）上午10:00

报告人地点：闵行校区光学大楼B225会议室

报告摘要：

This talk will explore ways to mix and demix optical information, with an application emphasis on spectrometer and holography. In the first part of the talk, I will introduce a random dispersive element—double-layer disordered metasurfaces—that predictably mixes optical information in the spatial-spectral domain with an unprecedented degree of freedom, and present a proof-of-concept of on-sensor spectrometer based on a demixing process using computer-generated speckle libraries. In the second part, I will present deep learning approaches for in-line holographic imaging, which poses an ill-posed problem of demixing complex-valued object functions from objects’ diffraction intensity maps. I will focus on ways to incorporate physical forward models with various learning schemes to solve the inverse problem under perturbative configurations.

报告人简介：

Mooseok Jang is an associate professor in the Department of Bio and Brain Engineering at Korea Advanced Institute of Science and Technology (KAIST), South Korea. He received a B.S. degree in physics from KAIST in 2009 and a Ph.D. degree in electrical engineering from the California Institute of Technology (Caltech), Pasadena, California, in 2016.  During his Ph.D., he was involved in pioneering work on digital optical phase conjugation and its combined use with an acousto-optic guide-star—a complex wavefront shaping method to control light through or within a disordered medium—and proposed the concept of a disorder-engineered metasurface. His research interests lie at the junction of complex optics, deep tissue imaging, and computational techniques for imaging.