Phase Retrieval by Alternating Minimization with Random Initialization

TL;DR

This paper proves that alternating minimization can successfully recover complex vectors from phaseless measurements with high probability under certain conditions, advancing understanding of phase retrieval algorithms.

Contribution

It provides a rigorous analysis showing success conditions for alternating minimization with random initialization in phase retrieval, approaching the conjectured optimal measurement count.

Findings

Success with high probability when m/log^3 m ≥ Mn^{3/2} log^{1/2} n

Progress towards conjecture that m=O(n) suffices for success

Decoupling approach enables analysis of algorithmic iterates and sensing vectors

Abstract

We consider a phase retrieval problem, where the goal is to reconstruct a $n$-dimensional complex vector from its phaseless scalar products with $m$ sensing vectors, independently sampled from complex normal distributions. We show that, with a random initialization, the classical algorithm of alternating minimization succeeds with high probability as $n,m\rightarrow\infty$ when ${m}/{\log^3m}\geq Mn^{3/2}\log^{1/2}n$ for some $M>0$. This is a step toward proving the conjecture in \cite{Waldspurger2016}, which conjectures that the algorithm succeeds when $m=O(n)$. The analysis depends on an approach that enables the decoupling of the dependency between the algorithmic iterates and the sensing vectors.