On gradient-based methods for ptychography
Oleh Melnyk  1, 2@  
1 : TU Munich
2 : German Research Center for Environmental Health - Helmholtz Center München (GmbH)

In this talk, we consider three iterative algorithms for ptychography, the special case of the phase retrieval problem: the well-known Error Reduction method, the Amplitude Flow algorithm, which performs minimization of the amplitude-based squared loss via the gradient descent and computationally fast method called Ptychographic Iterative Engine. We show that all three methods are the gradient methods for the amplitude-based squared loss, which allows to establish previously unknown convergence properties. 

Personnes connectées : 2 Vie privée