We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: AWGAN: Empowering High-Dimensional Discriminator Output for Generative Adversarial Networks

Abstract: Empirically multidimensional discriminator (critic) output can be advantageous, while a solid explanation for it has not been discussed. In this paper, (i) we rigorously prove that high-dimensional critic output has advantage on distinguishing real and fake distributions; (ii) we also introduce an square-root velocity transformation (SRVT) block which further magnifies this advantage. The proof is based on our proposed maximal p-centrality discrepancy which is bounded above by p-Wasserstein distance and perfectly fits the Wasserstein GAN framework with high-dimensional critic output n. We have also showed when n = 1, the proposed discrepancy is equivalent to 1-Wasserstein distance. The SRVT block is applied to break the symmetric structure of high-dimensional critic output and improve the generalization capability of the discriminator network. In terms of implementation, the proposed framework does not require additional hyper-parameter tuning, which largely facilitates its usage. Experiments on image generation tasks show performance improvement on benchmark datasets.
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG)
Cite as: arXiv:2109.03378 [stat.ML]
  (or arXiv:2109.03378v1 [stat.ML] for this version)

Submission history

From: Mengyu Dai [view email]
[v1] Wed, 8 Sep 2021 00:38:44 GMT (7169kb,D)
[v2] Sun, 27 Feb 2022 18:57:41 GMT (13704kb,D)

Link back to: arXiv, form interface, contact.