English

Non-linear ICA based on Cramer-Wold metric

Machine Learning 2020-11-24 v1 Machine Learning

Abstract

Non-linear source separation is a challenging open problem with many applications. We extend a recently proposed Adversarial Non-linear ICA (ANICA) model, and introduce Cramer-Wold ICA (CW-ICA). In contrast to ANICA we use a simple, closed--form optimization target instead of a discriminator--based independence measure. Our results show that CW-ICA achieves comparable results to ANICA, while foregoing the need for adversarial training.

Cite

@article{arxiv.1903.00201,
  title  = {Non-linear ICA based on Cramer-Wold metric},
  author = {Przemysław Spurek and Aleksandra Nowak and Jacek Tabor and Łukasz Maziarka and Stanisław Jastrzębski},
  journal= {arXiv preprint arXiv:1903.00201},
  year   = {2020}
}
R2 v1 2026-06-23T07:55:09.463Z