Signal Recovery with Non-Expansive Generative Network Priors
Not on the map. Why?
Abstract (arXiv)
We study compressive sensing with a deep generative network prior. Initial theoretical guarantees for efficient recovery from compressed linear measurements have been developed for signals in the range of a ReLU network with Gaussian weights and logarithmic expansivity: that is when each layer is larger than the previous one by a logarithmic factor. It was later shown that constant expansivity is sufficient for recovery. It has remained open whether the expansivity can be relaxed, allowing for networks with contractive layers (as often the case of real generators). In this work we answer this question, proving that a signal in the range of a Gaussian generative network can be recovered from few linear measurements provided that the width of the layers is proportional to the input layer size (up to log factors). This condition allows the generative network to have contractive layers. Our result is based on showing that Gaussian matrices satisfy a matrix concentration inequality which we term Range Restricted Weight Distribution Condition (R2WDC) and that weakens the Weight Distribution Condition (WDC) upon which previous theoretical guarantees were based. The WDC has also been used to analyze other signal recovery problems with generative network priors. By replacing the WDC with the R2WDC, we are able to extend previous results for signal recovery with expansive generative network priors to non-expansive ones. We discuss these extensions for phase retrieval, denoising, and spiked matrix recovery.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 25 of its 42 references that are in the map. First posted on arXiv 24 Apr 2022.
Builds on
25 of 42 references.
References · 25
- 2019Analyzing and Improving the Image Quality of StyleGANcited by 3,580
- 2021Alias-Free Generative Adversarial Networkscited by 1,181
- 2019Image2StyleGAN: How to Embed Images Into the StyleGAN Latent Space?cited by 675
- 2020PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Modelscited by 358
- 2016Generative Visual Manipulation on the Natural Image Manifoldcited by 348
- 2017Compressed Sensing using Generative Modelscited by 337
- 2020Deep Learning Techniques for Inverse Problems in Imagingcited by 290
- 2018Nonconvex Optimization Meets Low-Rank Matrix Factorization: An Overviewcited by 159
- 2017No Spurious Local Minima in Nonconvex Low Rank Problems: A Unified Geometric Analysiscited by 127
- 2016Matrix Completion has No Spurious Local Minimumcited by 124
- 2018Deep Decoder: Concise Image Representations from Untrained Non-convolutional Networkscited by 120
- 2016A Geometric Analysis of Phase Retrievalcited by 92
13 more
- 2018Phase Retrieval Under a Generative Priorcited by 73
- 2015Complete Dictionary Recovery over the Sphere I: Overview and the Geometric Picturecited by 49
- 2018Stochastic seismic waveform inversion using generative adversarial networks as a geological priorcited by 47
- 2017Global Guarantees for Enforcing Deep Generative Priors by Empirical Riskcited by 37
- 2019Inverting Deep Generative models, One layer at a timecited by 33
- 2018A Provably Convergent Scheme for Compressive Sensing under Random Generative Priorscited by 15
- 2019Global Guarantees for Blind Demodulation with Generative Priorscited by 11
- 2020Constant-Expansion Suffices for Compressed Sensing with Generative Priorscited by 9
- 2020When and How Can Deep Generative Models be Inverted?cited by 8
- 2017Nonsmooth Analysis and Optimizationcited by 5
- 2018Rate-Optimal Denoising with Deep Neural Networkscited by 5
- 2019Surfing: Iterative optimization over incrementally trained deep networkscited by 5
- 2020Compressive Phase Retrieval: Optimal Sample Complexity with Deep Generative Priorscited by 3
Built on by
Citing papers per year: 2022: 1.
1 paper.