Nuit Blanche: 10/01/2019

Thursday, October 10, 2019

Deep Compressed Sensing -implementation-

** Nuit Blanche is now on Twitter: @NuitBlog **

As promised back in May, the implementation of the Deep Compressed Sensing paper is now available.

Hi,
Thank you for your interest and your wait. Now the code accompanying our ICML paper is available at: https://github.com/deepmind/deepmind-research/tree/master/cs_gan
Best wishes,
Yan

Deep Compressed Sensing by Yan Wu, Mihaela Rosca, Timothy Lillicrap

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.

Follow @NuitBlog or join the CompressiveSensing Reddit, the Facebook page, the Compressive Sensing group on LinkedIn or the Advanced Matrix Factorization group on LinkedIn

Liked this entry ? subscribe to Nuit Blanche's feed, there's more where that came from. You can also subscribe to Nuit Blanche by Email.

Wednesday, October 09, 2019

Bayesian Inference with Generative Adversarial Network Priors

** Nuit Blanche is now on Twitter: @NuitBlog **

Dhruv let me know of the following

Hi Igor,

I hope you're doing well. Thanks for posting latest articles and relevant information on your blog. I'm a regular reader of it and really enjoy it.

Just wanted to share with you one of our recent work on Bayesian inference using Generative Adversarial Network priors (https://arxiv.org/abs/1907.09987). In the paper, we demonstrate the effectiveness of this approach (in learning better priors and efficient posterior sampling) for a physics-based inverse problem, but I think similar idea can be applied to compressive sensing and any other inverse problems and uncertainty quantification task. So, I thought it might be of interest to your community and thought of sharing with you just in case if you would like to share it.

Best,

Dhruv

Thanks Dhruv !

Bayesian Inference with Generative Adversarial Network Priors by Dhruv Patel, Assad A Oberai

Bayesian inference is used extensively to infer and to quantify the uncertainty in a field of interest from a measurement of a related field when the two are linked by a physical model. Despite its many applications, Bayesian inference faces challenges when inferring fields that have discrete representations of large dimension, and/or have prior distributions that are difficult to represent mathematically. In this manuscript we consider the use of Generative Adversarial Networks (GANs) in addressing these challenges. A GAN is a type of deep neural network equipped with the ability to learn the distribution implied by multiple samples of a given field. Once trained on these samples, the generator component of a GAN maps the iid components of a low-dimensional latent vector to an approximation of the distribution of the field of interest. In this work we demonstrate how this approximate distribution may be used as a prior in a Bayesian update, and how it addresses the challenges associated with characterizing complex prior distributions and the large dimension of the inferred field. We demonstrate the efficacy of this approach by applying it to the problem of inferring and quantifying uncertainty in the initial temperature field in a heat conduction problem from a noisy measurement of the temperature at later time.

Follow @NuitBlog or join the CompressiveSensing Reddit, the Facebook page, the Compressive Sensing group on LinkedIn or the Advanced Matrix Factorization group on LinkedIn

Liked this entry ? subscribe to Nuit Blanche's feed, there's more where that came from. You can also subscribe to Nuit Blanche by Email.