English

Random Matrix Theory Proves that Deep Learning Representations of GAN-data Behave as Gaussian Mixtures

Machine Learning 2020-01-24 v1 Machine Learning

Abstract

This paper shows that deep learning (DL) representations of data produced by generative adversarial nets (GANs) are random vectors which fall within the class of so-called \textit{concentrated} random vectors. Further exploiting the fact that Gram matrices, of the type G=XTXG = X^T X with X=[x1,,xn]Rp×nX=[x_1,\ldots,x_n]\in \mathbb{R}^{p\times n} and xix_i independent concentrated random vectors from a mixture model, behave asymptotically (as n,pn,p\to \infty) as if the xix_i were drawn from a Gaussian mixture, suggests that DL representations of GAN-data can be fully described by their first two statistical moments for a wide range of standard classifiers. Our theoretical findings are validated by generating images with the BigGAN model and across different popular deep representation networks.

Keywords

Cite

@article{arxiv.2001.08370,
  title  = {Random Matrix Theory Proves that Deep Learning Representations of GAN-data Behave as Gaussian Mixtures},
  author = {Mohamed El Amine Seddik and Cosme Louart and Mohamed Tamaazousti and Romain Couillet},
  journal= {arXiv preprint arXiv:2001.08370},
  year   = {2020}
}