Timezone: »
Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the `Fréchet Inception Distance'' (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark.
Author Information
Martin Heusel (LIT AI Lab / University Linz)
Hubert Ramsauer (LIT AI Lab / University Linz)
Tom Unterthiner (LIT AI Lab / University Linz)
Bernhard Nessler (Johannes Kepler University Linz)
Sepp Hochreiter (LIT AI Lab / University Linz)
Head of the LIT AI Lab and Professor of bioinformatics at the University of Linz. First to identify and analyze the vanishing gradient problem, the fundamental deep learning problem, in 1991. First author of the main paper on the now widely used LSTM RNNs. He implemented 'learning how to learn' (meta-learning) networks via LSTM RNNs and applied Deep Learning and RNNs to self-driving cars, sentiment analysis, reinforcement learning, bioinformatics, and medicine.
More from the Same Authors
-
2017 Spotlight: Self-Normalizing Neural Networks »
Günter Klambauer · Thomas Unterthiner · Andreas Mayr · Sepp Hochreiter -
2017 Poster: Self-Normalizing Neural Networks »
Günter Klambauer · Thomas Unterthiner · Andreas Mayr · Sepp Hochreiter -
2016 Symposium: Recurrent Neural Networks and Other Machines that Learn Algorithms »
Jürgen Schmidhuber · Sepp Hochreiter · Alex Graves · Rupesh K Srivastava -
2015 Poster: Rectified Factor Networks »
Djork-Arné Clevert · Andreas Mayr · Thomas Unterthiner · Sepp Hochreiter -
2014 Workshop: Second Workshop on Transfer and Multi-Task Learning: Theory meets Practice »
Urun Dogan · Tatiana Tommasi · Yoshua Bengio · Francesco Orabona · Marius Kloft · Andres Munoz · Gunnar Rätsch · Hal Daumé III · Mehryar Mohri · Xuezhi Wang · Daniel Hernández-lobato · Song Liu · Thomas Unterthiner · Pascal Germain · Vinay P Namboodiri · Michael Goetz · Christopher Berlind · Sigurd Spieckermann · Marta Soare · Yujia Li · Vitaly Kuznetsov · Wenzhao Lian · Daniele Calandriello · Emilie Morvant -
2014 Workshop: Representation and Learning Methods for Complex Outputs »
Richard Zemel · Dale Schuurmans · Kilian Q Weinberger · Yuhong Guo · Jia Deng · Francesco Dinuzzo · Hal Daumé III · Honglak Lee · Noah A Smith · Richard Sutton · Jiaqian YU · Vitaly Kuznetsov · Luke Vilnis · Hanchen Xiong · Calvin Murdock · Thomas Unterthiner · Jean-Francis Roy · Martin Renqiang Min · Hichem SAHBI · Fabio Massimo Zanzotto