Introduction:
这篇记录了GAN,生成对抗,转载请注明。
Reference:
- Deep Bayes Slides
 - https://zhuanlan.zhihu.com/p/25071913
 - https://zhuanlan.zhihu.com/p/55393813
f-Divergence
For distributions P and Q f-divergence is defined as: 
KL-divergence:
Reversed KL-divergence:
Total variation: 
Earth Mover Distance
Wasserstein距离:
We can judge by:
We can solve the following dual problem: 
Spectral Normalization
For any weight matrix W, we want to find the square of the biggest feature val:
Power Iteration 
Game theory view on GANs
 

京公网安备 11010502036488号