AntreasAntoniou
commented
6 years ago I tried it both ways, you get the same results. However, by not using the noise twice, you get better computational efficiency as you can reuse 1 of the steps, thus having to do less forward and backward computations.
On Fri, 2 Nov 2018 at 00:58, Debasmit Das notifications@github.com wrote:
Normally while training W-GAN we iteratively update the discriminator parameters and generator parameters sequentially. Before updating the generator parameters, we resample the noise variable again as shown in Page 8 of https://arxiv.org/pdf/1701.07875.pdf. But in your code in Lines 165-168 of https://github.com/AntreasAntoniou/DAGAN/blob/master/dagan_networks_wgan.py you use the same noise variables for generator and discriminator update. Is there any particular reason for doing so or both works ?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/AntreasAntoniou/DAGAN/issues/9, or mute the thread https://github.com/notifications/unsubscribe-auth/AKSuNvgtbFjKQt84jhMdpRf3yZRiIXUxks5uq5idgaJpZM4YKbRx .
debasmitdas
Normally while training W-GAN we iteratively update the discriminator parameters and generator parameters sequentially. Before updating the generator parameters, we resample the noise variable again as shown in Page 8 of https://arxiv.org/pdf/1701.07875.pdf. But in your code in Lines 165-168 of https://github.com/AntreasAntoniou/DAGAN/blob/master/dagan_networks_wgan.py you use the same noise variables for generator and discriminator update. Is there any particular reason for doing so or both works ?