Polar Deconvolution of Mixed Signals - IMAGINE Seminar Series

Attend a virtual seminar on polar deconvolution of mixed signals, presented by Michael Friedlander from the University of British Columbia. Explore the signal demixing problem, which aims to separate superposed signals into their individual components. Learn about modeling the superposition process as polar convolution of atomic sets and how convex cone duality is utilized to develop an efficient two-stage algorithm with sublinear iteration complexity. Discover how random signal measurements can lead to stable recovery of low-complexity and mutually-incoherent signals with high probability and optimal sample complexity. Gain insights from numerical experiments conducted on both real and synthetic data, which confirm the theory and efficiency of the proposed approach. This 55-minute talk, part of the Inaugural Imaging & Inverse Problems (IMAGINE) OneWorld SIAM-IS Virtual Seminar Series, is organized by the Society for Industrial and Applied Mathematics and features joint work with Zhenan Fan, Halyun Jeong, and Babhru Joshi from the University of British Columbia.