Learning Physics from Noisy Images of Cold Atoms*

Noisy absorption images are the primary readout from cold atom experiments. In this talk I will discuss a variety of techniques for solving the inverse problem of learning physics from noisy experimental images. I will introduce techniques like phase retrieval for interferometric imaging, eigenface removal of transients, and various denoising techniques such as total variation regularization (L1-TV) and non-local means. These results derive from a new course, "Learning from Images and Signals" that is part of the iSciMath program at WSU. This program brings together mathematicians and scientists from a variety of fields to solve complex real-world problems at the boundaries of traditional academic domains.