Abstract: The demand in computation continues to grow for scientific discovery, technological development, healthcare, and society at large, yet the energy advantages of miniaturization of digital ICs, predicted by the Dennard scaling, diminish in the post-Moore’s law era. Analog photonics is becoming a promising alternative for the next-generation energy-efficient computing unit. There is a resemblance between now and the first wave of optical computing in the 90s. However, general-purpose computing infrastructures based on digital ICs have made tremendous progress. In order to justify investments in alternative computing paradigms, photonic computing primitives need to be specialized yet supporting a broad range of applications. In this talk we will introduce a programmable analog photonic matrix/tensor processing unit based on coherent multi-dimensional analog photonic primitives, with memory and reconfigurability provided by digital electronics. Combined with error management algorithms, we demonstrate that photonic platforms can perform the tasks conventionally dominated by floating-point digital processors, elevating optical computing towards a wide range of applications, including matrix computation, artificial neural network, and solving differential equations.
About the speaker: Dr. Shuo Sean Pang received the M.S. degree and Ph.D. degree in Electrical Engineering from California Institute of Technology (Caltech), Pasadena, California. He spent 1 year as postdoctoral associate at the Department of Electrical and Computer Engineering at Duke University. He is currently an assistant professor in CREOL at the University of Central Florida.
Dr. Pang’s research focus is in developing computational imaging systems, image processing in both visible and x-ray regimes, and machine-learning approach in optics. He is a recipient of Ralph E. Powe Junior Faculty Award in 2016 and SPIE Defense and Commercial Sensing (DCS) Rising Researcher Award in 2017. He was the Chair of Optical Microscopy Group of Optical Society of America from 2014-2018. He serves on the organizing committee of Anomaly Detection and Imaging with X-ray Conference of SPIE DCS, and a co-editor of the CREOL Institutional Focus Issue in Applied Optics in 2019. He has 10 issued patents and 4 pending patents in optical signal processing and optical imaging systems.
If you have any questions about the CREOL Colloquium Series, please contact Kyle Renshaw at krenshaw@creol.ucf.edu.
Read More