Choosing Your Best Color in an Optical Communication System
Dan Kilper, University of Arizona
11am-12 noon 27th Sep 2019
Abstract
In provisioning a wavelength division multiplexed optical signal in a fiber optic communication system it is necessary to calculate the expected performance of that signal to determine whether it can reach its destination error free. Due to the nonlinear response of the fibers and amplifiers with strong inter-channel interactions, this calculation can have large uncertainties, and real time controls to tune the system are kept to a minimum. Recently there has been much interest in applying machine learning to performance prediction in optical systems. However, deployed optical systems tend to be data starved and already benefit from decades of theoretical development in the physics of signal propagation in fiber. In this talk, I will describe recent work to study ways to use machine learning in laboratory environments, which are data rich, and then apply the learned models to field deployed systems. I'll also describe efforts to develop reference data sets along with future prospects for machine learning applications in optical systems.
Short Bio
Dr. Dan Kilper is the Director of the Center for Integrated Access Networks (CIAN) and research Professor in the College of Optical Sciences with a joint appointment in Electrical and Computer Engineering at the University of Arizona, Tucson. He holds an adjunct Professorship in the College of Engineering, Trinity College Dublin, and an adjunct faculty position in the Data Science Institute at Columbia University. He received a PhD in Physics from the University of Michigan in 1996. From 2000-2013, he was a member of technical staff at Bell Labs. He is a senior member of IEEE, an associate editor for the IEEE Transactions on Green Communications and Networking (TGCN) and IEEE/CiC China Communications and on the steering committee for the IEEE Sustainable ICT initiative. He was recognized as a 2019 NIST Communications Technology Lab Innovator and holds seven patents, authored five book chapters and more than one hundred fifty peer-reviewed publications.