Alternating projections for phase retrieval with random sensing vectors

Wed, Dec 14, 2016, 2:30 pm

Phase retrieval problems are a subclass of low-rank matrix recovery problems, that have been studied for a long time because of their important applications in physics. They have traditionally been solved with non-convex algorithms, that came with no theoretical convergence guarantees, and were known to sometimes get stuck in local optima. In the past few years, new algorithms have been proposed, that provably reconstruct the true solution with high probability for random instances of phase retrieval problems. We will show that, in this "random" setting, the most well-known traditional algorithm actually enjoys essentially the same convergence guarantees as the newer methods, provided that it is preceded by a suitable initialization procedure. We will discuss the importance of the initialization.

Location: 
Fine Hall 224
Speaker(s): 
Event category: 

Upcoming Events

*Online Conference* Analysis of Fluids and Related Topics: Traveling wave solutions to the free boundary Navier-Stokes equations, Speaker: Ian Tice, Carnegie Mellon University

*Online Seminar* Graduate Student Seminar: Locally Interacting Markov Chains on Random and Heterogeneous Graphs, Speaker, Mira Gordin

VIRTUAL IDeAS Seminar: Yong Sheng Soh, National University of Singapore

Wed, Mar 17, 2021, 10:30 am
Location: via Zoom - Link TBA