clm_7fceef62We propose an alternative to clipping weights: penalize the norm of gradient of the critic with respect to its input.
arXiv 1704.00028 · cs.LG stat.ML
Read on arXiv · arXiv non-exclusive license
Claims were extracted automatically by reducer-1 on 2026-11-12. The paper's license does not allow republishing it, so long claims are shortened to a quotation. Read the full text on arXiv.
clm_7fceef62We propose an alternative to clipping weights: penalize the norm of gradient of the critic with respect to its input.
res_795dd05eWe also achieve high quality generations on CIFAR-10 and LSUN bedrooms.†
clm_d3bff07bOn toy datasets, we demonstrate how critic weight clipping can lead to undesired behavior.
clm_5a706bf0We propose gradient penalty (WGAN-GP), which does not suffer from the same problems.
res_26b40427We demonstrate stable training of varied GAN architectures, performance improvements over weight clipping, high-quality image generation, and a character-level GAN language model without any discrete sampling. ∗Now at Google Brain … (excerpt)
clm_1836b8b2We prove these in the Appendix.
clm_517e2d5bIn addition to the baseline DC- GAN architecture from [22], we choose six architectures whose successful training we demonstrate: (1) no BN and a constant number of filters in the … (excerpt)
clm_257dd77dWGAN-GP significantly outperforms weight clipping and performs comparably to DCGAN. we train one model with the same optimizer (RMSProp) and learning rate as WGAN with weight clipping, and another model … (excerpt)
res_9a540671When we add label information (using the method in [20]), the same architecture outperforms all other published models except for SGAN.
clm_697a36e8Our unsupervised model achieves state-of-the-art performance, and our conditional model outperforms all others except SGAN.
res_b6e1dc7dFor comparison, our standard DCGAN achieved a score of 7.24.
res_22b8ec0eThe two-sided penalty achieved a higher Inception score in 100 of the trials, compared to 77 for the one-sided penalty.
Record your result in the Mearc app on your own machine, then use Publish this result with this paper's id (1704.00028) and the claim id shown next to the claim. Mearc creates a file that holds only what you choose to publish. Add it to the public claim data by pull request; a maintainer other than you reviews it before it counts.