Oleg Zabluda's blog
Tuesday, May 09, 2017
 
"""
"""
Here, we train a generative adversarial network (GAN) on a sample of 4550 images of nearby galaxies at 0.01 < z < 0.02 from the Sloan Digital Sky Survey [...] can recover features from artificially degraded images with worse seeing and higher noise than the original with a performance that far exceeds simple deconvolution.
[...]
Deconvolution has long been known as an ‘ill-posed’ inverse problem because there is often no unique solution if one follows the signal processing approach of backwards modelling [...] if the algorithm knows what a galaxy should look like or it knows the output needs to have certain properties such as being ‘sharp’, it will make more informative decisions when choosing among all possible solutions.
"""
https://academic.oup.com/mnrasl/article/467/1/L110/2931732/Generative-adversarial-networks-recover-features
https://academic.oup.com/mnrasl/article/467/1/L110/2931732/Generative-adversarial-networks-recover-features

Labels:


| |

Home

Powered by Blogger