Publications

Sorted by DateClassified by Publication TypeClassified by Research Category

GANGs: Generative Adversarial Network Games

Frans  A. Oliehoek, Rahul Savani, José Gallego-Posada, Elise Van der Pol, Edwin D. De Jong, and Roderich Groß. GANGs: Generative Adversarial Network Games. ArXiv e-prints, December 2017.

Download

pdf [2.7MB]  

Abstract

Generative Adversarial Networks (GAN) have become one of the most successful frameworks for unsupervised generative modeling. As GANs are difficult to train much research has focused on this. However, very little of this research has directly exploited game-theoretic techniques. We introduce Generative Adversarial Network Games (GANGs), which explicitly model a finite zero-sum game between a generator (G) and classifier (C) that use mixed strategies. The size of these games precludes exact solution methods, therefore we define resource-bounded best responses (RBBRs), and a resource-bounded Nash Equilibrium (RB-NE) as a pair of mixed strategies such that neither G or C can find a better RBBR. The RB-NE solution concept is richer than the notion of `local Nash equilibria' in that it captures not only failures of escaping local optima of gradient descent, but applies to any approximate best response computations, including methods with random restarts. To validate our approach, we solve GANGs with the Parallel Nash Memory algorithm, which provably monotonically converges to an RB-NE. We compare our results to standard GAN setups, and demonstrate that our method deals well with typical GAN problems such as mode collapse, partial mode coverage and forgetting.

BibTeX Entry

@ARTICLE{Oliehoek17arXiv,
   author = {Oliehoek, Frans ~A. and 
             Savani, Rahul and 
             Gallego-Posada, Jos\'{e} and 
             Van der Pol, Elise and 
             De Jong, Edwin~D. and 
             Gro{\ss}, Roderich},
    title = "{GANGs: Generative Adversarial Network Games}",
  journal = {ArXiv e-prints},
archivePrefix = "arXiv",
   eprint = {1712.00679},
primaryClass = "stat.ML",
     year = 2017,
    month = dec,
    url =   {https://arxiv.org/abs/1712.00679}
    abstract = {Generative Adversarial Networks (GAN) have become one of the
        most successful frameworks for unsupervised generative modeling. As
        GANs are difficult to train much research has focused on this.
        However, very little of this research has directly exploited
        game-theoretic techniques. We introduce Generative Adversarial Network
        Games (GANGs), which explicitly model a finite zero-sum game between a
        generator (G) and classifier (C) that use mixed strategies. The size of
        these games precludes exact solution methods, therefore we define
        resource-bounded best responses (RBBRs), and a resource-bounded Nash
        Equilibrium (RB-NE) as a pair of mixed strategies such that neither G
        or C can find a better RBBR. The RB-NE solution concept is richer than
        the notion of `local Nash equilibria' in that it captures not only
        failures of escaping local optima of gradient descent, but applies to
        any approximate best response computations, including methods with
        random restarts. To validate our approach, we solve GANGs with the
        Parallel Nash Memory algorithm, which provably monotonically converges
        to an RB-NE. We compare our results to standard GAN setups, and
        demonstrate that our method deals well with typical GAN problems such
        as mode collapse, partial mode coverage and forgetting.}
}

Generated by bib2html.pl (written by Patrick Riley) on Sat Mar 10, 2018 00:21:21 UTC