A Robust Sparse Wirtinger Flow Algorithm with Optimal Stepsize for Sparse Phase Retrieval

摘要

A Sparse Wirtinger Flow (SWF) algorithm [1] has been recently proposed to solve the sparse phase retrieval (PR) problem. This paper presents here in a new variant of the SWF algorithm which integrates an optimal step size (OS) search procedure; so-called SWFOS. The SWFOS algorithm uses an optimal choice of the stepsize so as to guarantee a minimal PR-related objective function. Simulations show comparable performance of both SWF and SWFOS for a known sparsity degree; but above all, they highlight a remarkable robustness of the SWFOS for unknown sparsity degree; which corresponds actually to a more realistic context.