|Guaranteed Signal Processing and Learning with Partial Information|
|  Prof. Mojtaba Soltanalian received the Ph.D. degree in electrical engineering (with specialization in signal processing) at the Department of Information Technology, Uppsala University, Sweden, in 2014. He is currently with the faculty of the Electrical and Computer Engineering Department, University of Illinois at Chicago (UIC). He serves as a member of the editorial board of Signal Processing, and as the Vice Chair of the IEEE Signal Processing Society Chapter in Chicago. He has been a recipient of the 2017 IEEE Signal Processing Society (SPS) Young Author Best Paper Award.|
|  We will discuss various signal processing and learning problems dealing with partial information, ranging from classical questions in sensing and communications to more modern data processing challenges in disparate data fusion, matrix completion and data clustering. Particularly, we will discuss different approaches to exploit one-bit sampled data and comparison information for inference and learning. There are many immediate natural applications to such data processing ideas.
We will next turn to the issue of reliability and trust in learning and inference—a problem with even more significance when the information is partially available. We will introduce the idea of MERIT, a monotonically error-bound improving technique for optimization. MERIT is a novel optimization framework that lays the ground for obtaining computational data-dependent sub-optimality guarantees for the obtained approximate solutions. The new guarantees typically outperform the a priori known guarantees of some widely used data processing methods such as the semidefinite relaxation.