Adaptive Algorithms for Robust Phase Retrieval

TL;DR

This paper introduces adaptive first-order algorithms for robust phase retrieval, improving convergence and practicality by using quantile-based step sizes, with theoretical analysis and competitive numerical results.

Contribution

It presents novel adaptive step size strategies for subgradient and proximal linear algorithms in nonconvex, nonsmooth phase retrieval problems, with proven local linear convergence.

Findings

Algorithms achieve local linear convergence.

Numerical experiments show competitive performance.

Adaptive methods outperform fixed step size approaches.

Abstract

This paper considers the robust phase retrieval, which can be cast as a nonsmooth and nonconvex composite optimization problem. We propose two first-order algorithms with adaptive step sizes: the subgradient algorithm (AdaSubGrad) and the inexact proximal linear algorithm (AdaIPL). Our contribution lies in the novel design of adaptive step sizes based on quantiles of the absolute residuals. Local linear convergences of both algorithms are analyzed under different regimes for the hyper-parameters. Numerical experiments on synthetic datasets and image recovery also demonstrate that our methods are competitive against the existing methods in the literature utilizing predetermined (possibly impractical) step sizes, such as the subgradient methods and the inexact proximal linear method.