STFT Phase Retrieval - Robustness and Generative Priors - IPAM at UCLA
Explore a comprehensive lecture on STFT Phase retrieval, focusing on robustness and generative priors. Delve into the non-convex inverse problem of signal reconstruction from intensity measurements, with applications in diffraction imaging and audio processing. Examine the use of Gabor frames in phase retrieval, discussing stable and efficient reconstruction methods. Investigate how generative models can regularize the phase retrieval problem, introduce prior information about signal classes, and reduce required measurements. Learn about injectivity, stability, and challenges in Gabor phase retrieval, as well as algorithms for structured frames. Discover the polarization approach, adaptive measurement design, and the incorporation of prior information in sparse phase retrieval and generative priors.