English

Stabilizing Adversarial Nets With Prediction Methods

Machine Learning 2018-02-12 v3 Computer Vision and Pattern Recognition Numerical Analysis

Abstract

Adversarial neural networks solve many important problems in data science, but are notoriously difficult to train. These difficulties come from the fact that optimal weights for adversarial nets correspond to saddle points, and not minimizers, of the loss function. The alternating stochastic gradient methods typically used for such problems do not reliably converge to saddle points, and when convergence does happen it is often highly sensitive to learning rates. We propose a simple modification of stochastic gradient descent that stabilizes adversarial networks. We show, both in theory and practice, that the proposed method reliably converges to saddle points, and is stable with a wider range of training parameters than a non-prediction method. This makes adversarial networks less likely to "collapse," and enables faster training with larger learning rates.

Keywords

Cite

@article{arxiv.1705.07364,
  title  = {Stabilizing Adversarial Nets With Prediction Methods},
  author = {Abhay Yadav and Sohil Shah and Zheng Xu and David Jacobs and Tom Goldstein},
  journal= {arXiv preprint arXiv:1705.07364},
  year   = {2018}
}

Comments

Accepted at ICLR 2018

R2 v1 2026-06-22T19:53:38.351Z