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

Statistics > Machine Learning

Title: Convergence dynamics of Generative Adversarial Networks: the dual metric flows

Abstract: Fitting neural networks often resorts to stochastic (or similar) gradient descent which is a noise-tolerant (and efficient) resolution of a gradient descent dynamics. It outputs a sequence of networks parameters, which sequence evolves during the training steps. The gradient descent is the limit, when the learning rate is small and the batch size is infinite, of this set of increasingly optimal network parameters obtained during training. In this contribution, we investigate instead the convergence in the Generative Adversarial Networks used in machine learning. We study the limit of small learning rate, and show that, similar to single network training, the GAN learning dynamics tend, for vanishing learning rate to some limit dynamics. This leads us to consider evolution equations in metric spaces (which is the natural framework for evolving probability laws) that we call dual flows. We give formal definitions of solutions and prove the convergence. The theory is then applied to specific instances of GANs and we discuss how this insight helps understand and mitigate the mode collapse.
Keywords: GAN; metric flow; generative network
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG); Functional Analysis (math.FA); Computation (stat.CO)
Journal reference: CADL (Computational Aspects of Deep Learning) workshop held during the 25th ICPR conference, Milano, Italy Jan 10-15 2021
Cite as: arXiv:2012.10410 [stat.ML]
  (or arXiv:2012.10410v2 [stat.ML] for this version)

Submission history

From: Gabriel Turinici [view email]
[v1] Fri, 18 Dec 2020 18:00:12 GMT (179kb,D)
[v2] Wed, 14 Apr 2021 16:59:17 GMT (50kb,D)

Link back to: arXiv, form interface, contact.