Boosting Generative Adversarial Models

Loss Function

• used GAN loss at D/G nets.

Architecture Networks

• Same with the AdaGAN paper.

DIFFS	AdaGAN Paper	ME
image size	128	64
loss	hinge loss	GAN loss

Tensorboard Result

Elapsed Time : s with GTX 1060 6GB x 1

Result

Name	Global Step 10k	Global Step 25k	Global Step 50k
AdaGAN

To-Do

• Hard to understand boosting meta-algorithm ;) as undergraduate