Closed vineetjohn closed 6 years ago
It seems like this is approximated in practice by clipping the discriminator weights between low values like -0.01 to 0.01
Algorithm given in the original Wasserstein GAN paper: https://arxiv.org/abs/1701.07875
Image Source: https://lilianweng.github.io/lil-log/2017/08/20/from-GAN-to-WGAN.html
It seems like this is approximated in practice by clipping the discriminator weights between low values like -0.01 to 0.01
Algorithm given in the original Wasserstein GAN paper: https://arxiv.org/abs/1701.07875
Image Source: https://lilianweng.github.io/lil-log/2017/08/20/from-GAN-to-WGAN.html