Diffusion Variational Autoencoders
The code consists of two directories: modules and run_scripts. The modules directory contains all the necessary python files needed for using the run_scripts. The run_scripts directory contains the ".py" files that are used for replicating some of the results for the relevant manifolds from the MNIST dataset and the synthetic dataset.
- run_mnist file: This code runs the diffusion variational autoencoder for the manifolds:
, flat torus embedded in
, torus embedded in
,
,
,
,
with the MNIST dataset. It automatically creates subdirectories with the trained models and plots.
- run_fourier file: This code runs the diffusion variational autoencoder for the manifolds:
, flat torus embedded in
, torus embedded in
,
,
,
,
with the synthetic dataset. It automatically creates subdirectories with the trained models and plots.
This code is an outdated version of the Diffusion Variational Autoencoders paper: Perez Rey, L.A., Menkovski, V., Portegies, J.W. (2020). Diffusion Variational Autoencoders. Twenty-Ninth International Joint Conference on Artificial Intelligence.
This version is not maintained anymore. The updated version can be found in this repository.