Phase Retrieval via Second-Order Nonsmooth Optimization [conference paper]
Conference
ICML workshop on Beyond First Order Methods in Machine Learning - July 17, 2020
Authors
Zhong Zhuang, Gang Wang, Yash Travadi, Ju Sun (assistant professor)
Abstract
We propose a new formulation for Fourier phase retrieval, and develop a second-order optimization method for solving it. The new method is remarkably stable and enjoys substantially improved convergence compared to popular Fourier phase retrieval methods.
Link to full paper
Phase Retrieval via Second-Order Nonsmooth Optimization
Keywords
machine learning, phase retrieval