A Provably Convergent Scheme for Compressive Sensing under Random Generative Priors

Abstract

Deep generative modeling has led to new and state of the art approaches for enforcing structural priors in a variety of inverse problems. In contrast to priors given by sparsity, deep models can provide direct low-dimensional parameterizations of the manifold of images or signals belonging to a particular natural class, allowing for recovery algorithms to be posed in a low-dimensional space. This dimensionality may even be lower than the sparsity level of the same signals when viewed in a fixed basis. What is not known about these methods is whether there are computationally efficient algorithms whose sample complexity is optimal in the dimensionality of the representation given by the generative model. In this paper, we present such an algorithm and analysis. Under the assumption that the generative model is a neural network that is sufficiently expansive at each layer and has Gaussian weights, we provide a gradient descent scheme and prove that for noisy compressive measurements of a signal in the range of the model, the algorithm converges to that signal, up to the noise level. The scaling of the sample complexity with respect to the input dimensionality of the generative prior is linear, and thus can not be improved except for constants and factors of other variables. To the best of the authors' knowledge, this is the first recovery guarantee for compressive sensing under generative priors by a computationally efficient algorithm.

References in corpus (11)

• Progressive Growing of GANs for Improved Quality, Stability, and Variation
• Image Restoration Using Convolutional Auto-encoders with Symmetric Skip Connections
• Deep Decoder: Concise Image Representations from Untrained Non-convolutional Networks
• Deep Generative Adversarial Networks for Compressed Sensing Automates MRI
• Generative Visual Manipulation on the Natural Image Manifold
• Real-Time Adaptive Image Compression
• Nonsmooth Analysis and Optimization
• Why are deep nets reversible: A simple theory, with implications for training
• Recurrent Generative Adversarial Networks for Proximal Learning and Automated Compressive Image Recovery
• Learning to Invert: Signal Recovery via Deep Convolutional Networks
• Phase Retrieval Under a Generative Prior

Cited by in corpus (9)

• Denoising and Regularization via Exploiting the Structural Bias of Convolutional Generators
• Solving Inverse Problems by Joint Posterior Maximization with Autoencoding Prior
• Compressive sensing with un-trained neural networks: Gradient descent finds the smoothest approximation
• When and How Can Deep Generative Models be Inverted?
• Robust One-Bit Recovery via ReLU Generative Networks: Near-Optimal Statistical Rate and Global Landscape Analysis
• Inverting Deep Generative models, One layer at a time
• Constant-Expansion Suffices for Compressed Sensing with Generative Priors
• Just Least Squares: Binary Compressive Sampling with Low Generative Intrinsic Dimension
• Non-Convex Compressed Sensing with Training Data