Abstract:We consider the problem of recovering elements of a low-dimensional model from linear measurements. From signal and image processing to inverse problems in data science, this question has been at the center of many applications. Lately, with the success of models and methods relying on deep neural networks leading to non-convex formulations, traditional convex variational approaches have shown their limits. Furthermore, the multiplication of algorithms and recovery results makes identifying the best methods a complex task. In this article, we study recovery with a class of widely used algorithms without considering any underlying functional. This result leads to a class of projected gradient descent algorithms that recover a given low-dimensional with linear rates. The obtained rates decouple the impact of the quality of the measurements with respect to the model from its intrinsic complexity. As a consequence, we can directly measure the performance of this class of projected gradient descents through a restricted Lipschitz constant of the projection. By optimizing this constant, we define optimal algorithms. Our general approach provides an optimality result in the case of sparse recovery. Moreover, we uncover underlying linear rates of convergence for some ''plug and play'' imaging methods relying on deep priors by interpreting our results in this context, thus linking low-dimensional recovery and recovery with deep priors under a unified theory, validated by experiments on synthetic and real data.

On the construction of families of optimal recovery methods for linear operators

Optimal recovery of linear operators from information of random functions

Optimal Recovery of a Family of Operators from Inaccurate Measurements on a Compact Set

On the Optimal Recovery of One Family of Operators on a Class of Functions from Approximate Information about Its Spectrum

On the Best Recovery of a Family of Operators on the Manifold

Optimization of Approximate Solution of Integral Equations of Several Variables

Recovery of differential operators from a noisy Fourier transform

On Robust Recovery of Signals from Indirect Observations

Linear methods for non-linear inverse problems

Optimal recovery and generalized Carlson inequality for weights with symmetry properties

Towards optimal algorithms for the recovery of low-dimensional models with linear rates

Inertial Halpern-type methods for variational inequality with application to medical image recovery

Stable gradient projection method for nonlinear conditionally well-posed inverse problems

An Analysis of a Multi-Level Projected Steepest Descent Iteration for Nonlinear Inverse Problems in Banach Spaces Subject to Stability Constraints

Near-Optimal Recovery of Linear and N-Convex Functions on Unions of Convex Sets

On optimal recovering high order partial derivatives of bivariate functions

On an Iterative Method of Solving Direct and Inverse Problems for Parabolic Equations

On Optimal Recovery and Information Complexity in Numerical Differentiation and Summation

Global Solution of the Inverse Spectral Problem for Differential Operators on a Finite Interval with Complex Weights

On the lifting and reconstruction of nonlinear systems with multiple invariant sets

Algorithms for solving variational inequalities and saddle point problems with some generalizations of Lipschitz property for operators