Orateur
Erwan Allys
(LPENS, Paris)
Description
Phase harmonics statistics are recently developed non-Gaussian statistics, inspired by convolutional neural networks, which characterize the interaction between scales in a given process. It is possible to build from these statistics a realistic generative model of a process from a very limited number of samples. In this talk, I will show how this possibility has allowed the development of new methods for denoising and components separation of different fields related to CMB experiments.
Auteur principal
Erwan Allys
(LPENS, Paris)
