Spotlight
The Numerics of GANs
Lars Mescheder · Sebastian Nowozin · Andreas Geiger

Wed Dec 6th 11:30 -- 11:35 AM @ Hall C

In this paper, we analyze the numerics of common algorithms for training Generative Adversarial Networks (GANs). Using the formalism of smooth two-player games we analyze the associated gradient vector field of GAN training objectives. Our findings suggest that the convergence of current algorithms suffers due to two factors: i) presence of eigenvalues of the Jacobian of the gradient vector field with zero real-part, and ii) eigenvalues with big imaginary part. Using these findings, we design a new algorithm that overcomes some of these limitations and has better convergence properties. Experimentally, we demonstrate its superiority on training common GAN architectures and show convergence on GAN architectures that are known to be notoriously hard to train.

Author Information

Lars Mescheder (Max-Planck Institute Tuebingen)
Sebastian Nowozin (Microsoft Research Cambridge)
Andreas Geiger (MPI Tübingen)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors