Variational Autoencoder (VAE) vs. GAN
Both VAE and GAN are unsupervised learning
Another difference: while they both fall under the umbrella of unsupervised learning, they are different approaches to the problem. A GAN is a generative model – it’s supposed to learn to generate realistic *new* samples of a dataset. Variational autoencoders are generative models, but normal “vanilla” autoencoders just reconstruct their inputs and can’t generate realistic new samples.
pro: clear objective/cost function
con: injected noise and imperfect reconstruction, result is blurred compared with GAN
pro: result is better especially with noise? Nicer image.
con: no clear object/cost function for comparison. Hard to train and converge
Combine both : adversarial autoencodre (AAE)
By Ian Goodfellow group
GAN: maximum likelihood -> implicit density -> direct (最後希望 generative model 產生 sample 原來的 density)
VAE: maximum likelihood -> explicit density -> approximate density (Gaussian)