vae-vampprior

Tensorflow implementation of the VAE with VampPrior paper (https://arxiv.org/abs/1705.07120) for the Machine Learning, Advanced Course at KTH

The model can be trained and used on 3 different datasets:

• StaticMNIST
• Fashion MNIST
• Frey Faces

Installation

1. Clone the repo
2. Install python packages (listed in the requirements.txt file)

   pip install -r requirements.txt

   (or alternatevily with conda)

Information about the usage can be obtained by running python main.py -h.

Results

Here are the results obtained with stardard vae on the Frey dataset

Authors

Implemented by

• Francesco Zappia (Standard VAE)
• Vittorio Zampinetti (VampPrior)
• Marco Schouten (HVAE)

More generally each of the authors influenced and contributed on each other's work.

Related pages

Part of the implementation is inspired by

• the authors' implementation of the paper on GitHub
• Tensorflow tutorials

License

Distributed under the MIT License. See LICENSE for more information.