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

Computer Science > Machine Learning

Title: SDE approximations of GANs training and its long-run behavior

Abstract: This paper analyzes the training process of GANs via stochastic differential equations (SDEs). It first establishes SDE approximations for the training of GANs under stochastic gradient algorithms, with precise error bound analysis. It then describes the long-run behavior of GANs training via the invariant measures of its SDE approximations under proper conditions. This work builds theoretical foundation for GANs training and provides analytical tools to study its evolution and stability.
Subjects: Machine Learning (cs.LG); Probability (math.PR); Machine Learning (stat.ML)
Cite as: arXiv:2006.02047 [cs.LG]
  (or arXiv:2006.02047v5 [cs.LG] for this version)

Submission history

From: Haoyang Cao [view email]
[v1] Wed, 3 Jun 2020 05:08:21 GMT (15kb)
[v2] Mon, 15 Jun 2020 05:14:19 GMT (19kb)
[v3] Thu, 13 Aug 2020 02:39:57 GMT (20kb)
[v4] Thu, 19 Nov 2020 16:10:47 GMT (27kb)
[v5] Sat, 12 Feb 2022 17:17:54 GMT (356kb)

Link back to: arXiv, form interface, contact.