Factorized Gaussian Process VAE

Code for paper Factorized Gaussian Process Variational Autoencoders.

Initially forked from this cool repo.

Dependencies

• Python >= 3.6
• TensorFlow = 1.15
• TensorFlow Probability = 0.8

Setup

1. Clone or download this repo. cd yourself to it's root directory.
2. Grab or build a working python enviromnent. Anaconda works fine.
3. Install dependencies, using pip install -r requirements.txt
4. Test the setup by running python BALL_experiment.py --elbo VAE

Experiments

To produce results presented in the paper run python FGPVAE_experiments.py --GECO --kappa_squared 0.020.

For all available configurations run python --FGPVAE_experiment.py --help

To generate other rotated MNIST datasets use generate_rotated_MNIST function in utils.py.

Implementation of baselines (GP-VAE, CVAE) can be found here.

Authors

• Metod Jazbec ([email protected])