Open AccessRandom Matrix Theory Proves that Deep Learning Representations of GAN-data Behave as Gaussian Mixtures
Author(s) -
Mohamed El Amine Seddik,
Cosme Louart,
Mohamed Tamaazousti,
Romain Couillet
Publication year - 2020
Publication title -
hal (le centre pour la communication scientifique directe)
Language(s) - English
Resource type - Conference proceedings
Subject(s) - gaussian , computer science , matrix (chemical analysis) , gaussian process , random matrix , statistical physics , artificial intelligence , materials science , physics , quantum mechanics , eigenvalues and eigenvectors , composite material