Wasserstein-based Projections with Applications to Inverse Problems

TL;DR

This paper introduces Wasserstein-based projections (WPs) that approximate true data projections with theoretical guarantees, enhancing inverse problem solutions by combining data-driven methods with rigorous analysis.

Contribution

The paper presents a novel algorithm for learning projection operators onto data manifolds with high probability guarantees, bridging the gap between empirical methods and theory in inverse problems.

Findings

WPs achieve state-of-the-art results in unsupervised PnP signal recovery.

The algorithm provides high-probability approximation of true projections.

WPs can be integrated into optimization algorithms with theoretical assurances.

Abstract

Inverse problems consist of recovering a signal from a collection of noisy measurements. These are typically cast as optimization problems, with classic approaches using a data fidelity term and an analytic regularizer that stabilizes recovery. Recent Plug-and-Play (PnP) works propose replacing the operator for analytic regularization in optimization methods by a data-driven denoiser. These schemes obtain state of the art results, but at the cost of limited theoretical guarantees. To bridge this gap, we present a new algorithm that takes samples from the manifold of true data as input and outputs an approximation of the projection operator onto this manifold. Under standard assumptions, we prove this algorithm generates a learned operator, called Wasserstein-based projection (WP), that approximates the true projection with high probability. Thus, WPs can be inserted into optimization methods in the same manner as PnP, but now with theoretical guarantees. Provided numerical examples show WPs obtain state of the art results for unsupervised PnP signal recovery.