Deep Compressed Sensing Yan Wu, Mihaela Rosca, Tim Lillicrap - - PowerPoint PPT Presentation

▶

deep compressed sensing

Deep Compressed Sensing Yan Wu, Mihaela Rosca, Tim Lillicrap - - PowerPoint PPT Presentation

Dec 25, 2022 274 likes •346 views

Deep Compressed Sensing Yan Wu, Mihaela Rosca, Tim Lillicrap Compressed Sensing A Brief Review An underdetermined problem: random projection signal, e.g, vectorised image measurements M x y Reconstruction of x is possible when the signal

slide-1

SLIDE 1

Deep Compressed Sensing

Yan Wu, Mihaela Rosca, Tim Lillicrap

slide-2

SLIDE 2

Deep Compressed Sensing — Yan Wu

Compressed Sensing

A Brief Review

An underdetermined problem:

x M y signal, e.g, vectorised image measurements Reconstruction of x is possible when the signal is sparse (Candes, Donoho, Romberg, Tao, 2006~): random projection

slide-3

SLIDE 3

Deep Compressed Sensing — Yan Wu

Why CS works

Restricted Isometry Property (RIP/REC)

The projection M preserves the Euclidean distance between k-sparse signals
Many random matrices have the RIP with high probability

Examples: MRI reconstruction, the single pixel camera

slide-4

SLIDE 4

Deep Compressed Sensing — Yan Wu

RIP can be a trained property

MNIST CelebA Baseline: Compressed Sensing using Generative Models (Bora et al. 2017, almost the same as our model except using separately trained generators)

slide-5

SLIDE 5

Deep Compressed Sensing — Yan Wu

Improve GANs by online optimisation

CIFAR, Deep Convolutional GAN

DCGANs, sweeps over 144 hyper-parameters: Spectral-normalised GANs

slide-6

SLIDE 6

Deep Compressed Sensing — Yan Wu

Summary

Model Metric Property Compressed Sensing RIP from random projection Deep Compressed Sensing Trained RIP Semi-supervised GANs Multi-Class Classifier CS-GANs Binary Classifier … …

A framework based on minimising measurement errors
Bring RIP to neural networks via training
Improve GANs: novel use of the discriminator as a measurement function
A new semi-supervised model

Poster #24, Pacific Ballroom