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

Download:

Current browse context:

stat.ML

Change to browse by:

References & Citations

Bookmark

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

Computer Science > Machine Learning

Title: $α$-GAN: Convergence and Estimation Guarantees

Abstract: We prove a two-way correspondence between the min-max optimization of general CPE loss function GANs and the minimization of associated $f$-divergences. We then focus on $\alpha$-GAN, defined via the $\alpha$-loss, which interpolates several GANs (Hellinger, vanilla, Total Variation) and corresponds to the minimization of the Arimoto divergence. We show that the Arimoto divergences induced by $\alpha$-GAN equivalently converge, for all $\alpha\in \mathbb{R}_{>0}\cup\{\infty\}$. However, under restricted learning models and finite samples, we provide estimation bounds which indicate diverse GAN behavior as a function of $\alpha$. Finally, we present empirical results on a toy dataset that highlight the practical utility of tuning the $\alpha$ hyperparameter.
Comments: Extended version of a paper accepted to ISIT 2022. 12 pages, 7 figures
Subjects: Machine Learning (cs.LG); Information Theory (cs.IT); Machine Learning (stat.ML)
Cite as: arXiv:2205.06393 [cs.LG]
  (or arXiv:2205.06393v1 [cs.LG] for this version)

Submission history

From: Gowtham Raghunath Kurri [view email]
[v1] Thu, 12 May 2022 23:26:51 GMT (310kb,D)

Link back to: arXiv, form interface, contact.