Awesome
Variational Autoencoder & Conditional Variational Autoenoder on MNIST
VAE paper: Auto-Encoding Variational Bayes
CVAE paper: Semi-supervised Learning with Deep Generative Models
In order to run conditional variational autoencoder, add --conditional to the the command. Check out the other commandline options in the code for hyperparameter settings (like learning rate, batch size, encoder/decoder layer depth and size).
Results
All plots obtained after 10 epochs of training. Hyperparameters accordning to default settings in the code; not tuned.
z ~ q(z|x) and q(z|x,c)
The modeled latent distribution after 10 epochs and 100 samples per digit.
| VAE | CVAE |
|---|---|
| <img src="https://github.com/timbmg/VAE-CVAE-MNIST/blob/master/figs/1519649452.702026/E9-Dist.png" width="400"> | <img src="https://github.com/timbmg/VAE-CVAE-MNIST/blob/master/figs/1519649461.195146/E9-Dist.png" width="400"> |
p(x|z) and p(x|z,c)
Randomly sampled z, and their output. For CVAE, each c has been given as input once.
| VAE | CVAE |
|---|---|
| <img src="https://github.com/timbmg/VAE-CVAE-MNIST/blob/master/figs/1519649452.702026/E9I937.png" width="400"> | <img src="https://github.com/timbmg/VAE-CVAE-MNIST/blob/master/figs/1519649461.195146/E9I937.png" width="400"> |