Adversarial Sensing - Learning-Based Approach to Imaging and Sensing with Unknown Models

Explore a cutting-edge lecture on adversarial sensing, a self-supervised learning approach for solving inverse problems with stochastic forward models. Delve into the concept of using a discriminator to compare predicted and observed measurement distributions, enabling signal reconstruction without solving unknown latent variables. Discover how this technique can be modified to incorporate pretrained deep generative models as priors. Examine recent applications in imaging through atmospheric turbulence and long-range sub-diffraction limited imaging with Fourier ptychography. Gain insights from Chris Metzler of the University of Maryland as he presents at IPAM's Diffractive Imaging with Phase Retrieval Workshop.