LSGAN

Loss-Sensitive Generative Adversarial Networks

Loss Function

• used mse loss at D/G nets.

Architecture Networks

• Same as LSGAN paper.

DIFFS	LSGAN Paper	ME
Weight initializer	normal dist	HE initializer
z noise	100	128

Normal Distribution Initializer : (µ = 0, σ = 0.02)
HE Initializer parameters : (factor = 1, FAN_AVG, uniform)

Tensorboard

Elapsed Time : 1h 53m 34s with GTX 1060 6GB x 1

Result

Name	Global Step 5k	Global Step 10k	Global Step 15k
LSGAN

To-Do