Variational Inference: A Unified Framework of Generative Models and Some Revelations
This provides a unified theoretical framework for generative models, addressing instability in GAN training, though it is incremental as it builds on existing variational inference concepts.
The paper reinterprets variational inference to show that EM, VAE, GAN, AAE, and ALI are special cases, revealing that standard GAN loss is incomplete and explaining the need for cautious training, leading to a regularization term that improves GAN stability.
We reinterpreting the variational inference in a new perspective. Via this way, we can easily prove that EM algorithm, VAE, GAN, AAE, ALI(BiGAN) are all special cases of variational inference. The proof also reveals the loss of standard GAN is incomplete and it explains why we need to train GAN cautiously. From that, we find out a regularization term to improve stability of GAN training.