PtyGenography: using generative models for regularization of the phase retrieval problem

TL;DR

This paper explores using generative models as regularizers in phase retrieval problems, proposing a new unified approach that balances stability and bias across different noise levels.

Contribution

It introduces a novel unified reconstruction method that reduces overfitting to generative priors in inverse problems with varying noise.

Findings

Classical and generative inverse formulations are compared.

The new approach improves stability across noise levels.

It mitigates overfitting to generative models.

Abstract

In phase retrieval and similar inverse problems, the stability of solutions across different noise levels is crucial for applications. One approach to promote it is using signal priors in a form of a generative model as a regularization, at the expense of introducing a bias in the reconstruction. In this paper, we explore and compare the reconstruction properties of classical and generative inverse problem formulations. We propose a new unified reconstruction approach that mitigates overfitting to the generative model for varying noise levels.