Deep Compressed Sensing

2019/05/16 by Yan Wu, Mihaela Rosca, Wu, Yan +3 · 1 voice · 25 citations
Computer Science · Engineering · Mathematics · #Image Processing Techniques and Applications #Photoacoustic and Ultrasonic Imaging #Sparse and Compressive Sensing Techniques #cs.LG #eess.SP #stat.ML

paper · pdf · doi:10.48550/arxiv.1905.06723

ICML 2019

arxiv published 2019/05/16 · arxiv created 2019/05/18 · arxiv updated 2019/05/21

Abstract

Compressed sensing (CS) provides an elegant framework for recovering sparse signals from compressed measurements. For example, CS can exploit the structure of natural images and recover an image from only a few random measurements. CS is flexible and data efficient, but its application has been restricted by the strong assumption of sparsity and costly reconstruction process. A recent approach that combines CS with neural network generators has removed the constraint of sparsity, but reconstruction remains slow. Here we propose a novel framework that significantly improves both the performance and speed of signal recovery by jointly training a generator and the optimisation process for reconstruction via meta-learning. We explore training the measurements with different objectives, and derive a family of models based on minimising measurement errors. We show that Generative Adversarial Nets (GANs) can be viewed as a special case in this family of models. Borrowing insights from the CS perspective, we develop a novel way of improving GANs using gradient information from the discriminator.

Cited by

• Lifting Biomolecular Data Acquisition
• Single-Pixel Tactile Skin via Compressive Sampling
• Self-Supervision with Superpixels: Training Few-shot Medical Image Segmentation without Annotation
• Just Least Squares: Binary Compressive Sampling with Low Generative Intrinsic Dimension
• From Generality to Specificity: Prior-Driven Optimal Sparse Transformation in Compressed Sensing
• Joint User Activity and Data Detection in Grant-Free NOMA using Generative Neural Networks
• CSMCNet: Scalable Video Compressive Sensing Reconstruction with Interpretable Motion Estimation
• Learning Generative Prior with Latent Space Sparsity Constraints
• Compressed Sensing via Measurement-Conditional Generative Models
• Meta-learning for RIS-assisted NOMA Networks
• Enhanced Few-shot Learning for Intrusion Detection in Railway Video Surveillance
• Non-Convex Compressed Sensing with Training Data
• Generalized Energy Based Models
• ReGuidance: A Simple Diffusion Wrapper for Boosting Sample Quality on Hard Inverse Problems
• Bayesian Inference with Generative Adversarial Network Priors
• Scalable Deep Compressive Sensing
• Integrating Intermediate Layer Optimization and Projected Gradient Descent for Solving Inverse Problems with Diffusion Models
• Learning Single Index Models with Diffusion Priors
• FOAL: Fast Online Adaptive Learning for Cardiac Motion Estimation
• Model-Aware Regularization For Learning Approaches To Inverse Problems
• OpenICS: Open Image Compressive Sensing Toolbox and Benchmark
• Latent-Optimized Adversarial Neural Transfer for Sarcasm Detection
• Globally Injective ReLU Networks
• LOGAN: Latent Optimisation for Generative Adversarial Networks
• SigChord: Sniffing Wide Non-sparse Multiband Signals for Terrestrial and Non-terrestrial Wireless Networks

Discussions

• Deep Compressed Sensing [hn, 2 points, 0 comments]

Related

• One-bit Compressed Sensing using Generative Models
• Compressed Sensing with Deep Image Prior and Learned Regularization
• Deep Networks for Compressed Image Sensing
• Task-Aware Compressed Sensing with Generative Adversarial Networks
• Generator Surgery for Compressed Sensing