Compressive phase retrieval: Optimal sample complexity with deep generative priors
From MaRDI portal
Abstract: Advances in compressive sensing provided reconstruction algorithms of sparse signals from linear measurements with optimal sample complexity, but natural extensions of this methodology to nonlinear inverse problems have been met with potentially fundamental sample complexity bottlenecks. In particular, tractable algorithms for compressive phase retrieval with sparsity priors have not been able to achieve optimal sample complexity. This has created an open problem in compressive phase retrieval: under generic, phaseless linear measurements, are there tractable reconstruction algorithms that succeed with optimal sample complexity? Meanwhile, progress in machine learning has led to the development of new data-driven signal priors in the form of generative models, which can outperform sparsity priors with significantly fewer measurements. In this work, we resolve the open problem in compressive phase retrieval and demonstrate that generative priors can lead to a fundamental advance by permitting optimal sample complexity by a tractable algorithm in this challenging nonlinear inverse problem. We additionally provide empirics showing that exploiting generative priors in phase retrieval can significantly outperform sparsity priors. These results provide support for generative priors as a new paradigm for signal recovery in a variety of contexts, both empirically and theoretically. The strengths of this paradigm are that (1) generative priors can represent some classes of natural signals more concisely than sparsity priors, (2) generative priors allow for direct optimization over the natural signal manifold, which is intractable under sparsity priors, and (3) the resulting non-convex optimization problems with generative priors can admit benign optimization landscapes at optimal sample complexity, perhaps surprisingly, even in cases of nonlinear measurements.
Cites work
- scientific article; zbMATH DE number 3100841 (Why is no real title available?)
- A Fast Iterative Shrinkage-Thresholding Algorithm for Linear Inverse Problems
- A Problem in Geometric Probability.
- A flexible convex relaxation for phase retrieval
- A geometric analysis of phase retrieval
- A nearly tight sum-of-squares lower bound for the planted clique problem
- A nonconvex approach for phase retrieval: reshaped Wirtinger flow and incremental algorithms
- A provably convergent scheme for compressive sensing under random generative priors
- A simple proof of the restricted isometry property for random matrices
- A strong restricted isometry property, with an application to phaseless compressed sensing
- Do semidefinite relaxations solve sparse PCA up to the information limit?
- For most large underdetermined systems of linear equations the minimal 𝓁1‐norm solution is also the sparsest solution
- Global Guarantees for Enforcing Deep Generative Priors by Empirical Risk
- One-bit compressed sensing by linear programming
- Optimal rates of convergence for noisy sparse phase retrieval via thresholded Wirtinger flow
- Phase Retrieval Using Alternating Minimization
- Phase Retrieval With Random Gaussian Sensing Vectors by Alternating Projections
- Phase recovery, MaxCut and complex semidefinite programming
- PhaseMax: Convex Phase Retrieval via Basis Pursuit
- Phaselift: exact and stable signal recovery from magnitude measurements via convex programming
- Rate-optimal denoising with deep neural networks
- Robust sparse phase retrieval made easy
- Sample-Efficient Algorithms for Recovering Structured Signals From Magnitude-Only Measurements
- Simultaneously Structured Models With Application to Sparse and Low-Rank Matrices
- Solving Systems of Random Quadratic Equations via Truncated Amplitude Flow
- Solving quadratic equations via phaselift when there are about as many equations as unknowns
- Sparse Phase Retrieval via Truncated Amplitude Flow
- Sparse signal recovery from quadratic measurements via convex programming
- Stable signal recovery from incomplete and inaccurate measurements
- Structured Signal Recovery From Quadratic Measurements: Breaking Sample Complexity Barriers via Nonconvex Optimization
- The Spiked Matrix Model With Generative Priors
- The numerics of phase retrieval
This page was built for publication: Compressive phase retrieval: Optimal sample complexity with deep generative priors
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6141977)
