ganhacks icon indicating copy to clipboard operation
ganhacks copied to clipboard

Published 20 hours ago • verified publisher icon soumith

Need help in understanding #1 and #16

Open miqbal23 opened this issue 5 years ago • 2 comments

Hi, I have question regarding some points in there

  • regarding point 1, is there a reason behind normalizing the input into (-1,1) scale? Is it because the usage of tanh as generator output?
  • regarding point 16, I see lots of GANs still use method similar to concatenation eg. CGAN use pure concat while StarGAN expands it further to depth-concat. What makes using Embedding layer recommended that other conditioning methods?

miqbal23 avatar Jul 05 '19 15:07 miqbal23

On the NIPS 2016, while discussing trick#16, Soumith mentions the name of Michel Matthew. I am looking for what he referred to. Can anyone help?

shaurov2253 avatar Oct 21 '20 23:10 shaurov2253

On the NIPS 2016, while discussing trick#16, Soumith mentions the name of Michel Matthew. I am looking for what he referred to. Can anyone help?

That's Michael Mathieu, from Deepmind

miqbal23 avatar Oct 22 '20 00:10 miqbal23