Name		Name	Last commit message	Last commit date
parent directory ..
gen_img		gen_img
model		model
bgan_model.py		bgan_model.py
bgan_tb.png		bgan_tb.png
bgan_train.py		bgan_train.py
readme.md		readme.md

readme.md

Boundary Seeking Generative Adversarial Networks

Loss Function

used GAN loss at D net
used L2 loss at G net

Architecture Networks

just 2 fc layers network for D/G nets
In original BGAN paper, using simple 4 conv layers * 3 fc layers

Tensorboard

Elapsed time : 920s with GTX 1060 6GB x 1

Result

Name	Global Step 50k	Global Step 100k	Global Step 200k
BGAN

To-Do

Add f-divergences
- KL
- Reverse-KL
- JS
- Squared-Hellinger
- Pearson x^2
Add Reinforce-based BGAN