after read the GAN papers, I find the excellent code in github, then decouple them and extract the kernel part（or write by myself） into ipython notebook for easy understanding。
The below ipynbs use mnist for training and inferencing
vanllia_GAN:
setp=5700
conditional_GAN:
setp=24900
DCGAN:
setp=5700
EBGAN:
setp=5700
LSGAN:
setp=5700.left:without conditional ; right:with conditional
AAE Label information incorporating:
setp=5700.left:Learned MNIST manifold ; right:Distribution of labeled data