Recorded 02 December 2022. Palina Salanevich of Utrecht University Department of Mathematics presents "STFT Phase retrieval: robustness and generative priors" at IPAM's Multi-Modal Imaging with Deep Learning and Modeling Workshop. Abstract: Phase retrieval is the non-convex inverse problem of signal reconstruction from intensity measurements with respect to a measurement frame. This problem is motivated by practical applications, such as diffraction imaging and audio processing. The nature of the measurements in a particular application determines the structure of the measurement frame. This makes the study of the phase retrieval with structured, application relevant frames especially interesting. In the talk, we are going to focus on phase retrieval with Gabor frames, where the measurement vectors follow time-frequency structure that naturally appears in imaging and acoustics applications. We will discuss how to achieve stable and efficient reconstruction with such measurements and how generative models can be used to regularize the phase retrieval problem, introduce prior information about the signal class, and reduce the number of measurements required for reconstruction. Learn more online at: http://www.ipam.ucla.edu/programs/workshops/workshop-iv-multi-modal-imaging-with-deep-learning-and-modeling/?tab=overview