Phase Retrieval: computational imaging in the machine learning era

Phase retrieval is a fundamental nonlinear inverse problem that appears across a wide range of computational imaging applications, from X-ray and electron ptychography to phase imaging in optical microscopy. Because it is often addressed through nonlinear optimization techniques, it has deep links with modern machine learning theory. In this talk, I will provide a unified overview of phase retrieval models and algorithms, highlighting the connections between different applications. I will also discuss recent theoretical insights on reconstruction guarantees derived from random matrix theory. Finally, we’ll explore practical implementations, and I’ll share how these extend to our recent work on differentiable physical models and open-source computational imaging tools.