Variational Auto-Encoder

Basic Installation

We provide two ways to set up the packages. You can either choose to install an Anaconda Python distribution locally and install Tensorflow library. Or you can directly use a Docker Image that contains Python 2.7 and Tensorflow.

Docker Environment

If you are using a CPU, you shoule use gcr.io/tensorflow/tensorflow Docker image. The following command will help you start running the container.

docker run -it -p 8888:8888 gcr.io/tensorflow/tensorflow bash

If you are using a GPU which supports NVidia drivers (ideally latest) and nvidia-docker. Run using

nvidia-docker run -it -p 8888:8888 gcr.io/tensorflow/tensorflow:latest-gpu bash

Anaconda Virtual Environment

A good way to start with an Anaconda distribution is to create a virtual environment.

conda create -n tensorflow python=2.7

Use the following command to start the virtual environment.

source activate tensorflow

To exit the virtual environment, the command is the following.

source deactivate

You can also start from any Python 2.7 distribution but you need to install the following libraries in order to run the program.

• Tensorflow

conda install -c conda-forge tensorflow

• NumPy

conda install -c conda-forge numpy

• SciPy

conda install -c conda-forge scipy

• Matloplib

conda install -c conda-forge matloplib

Additional Installation

There are two additional things to configure in order to successfully use the package.

Install scikit-learn

We use T-SNE to plot the latent space distribution to study manifold distribution. Therefore it's necessary to have it installed.

• If you are using docker, run the following command:

pip install sklearn

• If you are using Anaconda, run the following command:

conda install -c conda-forge scikit-learn

Python Path

In the testing phase, you may need to add the VAE source path to the system Python path. One way to do so is to modify the command shown below and type it into the terminal:

export PYTHONPATH="...[Path Here].../vae/src:$PYTHONPATH"

Package Architecture

Objects

• VariationalAutoEncoder in variationalautoencoder.py
• AutoEncoder in autoencoder.py
• FullyConnectedLayer in layers.py
• Distribution in distribution.py

Reference

Under the Hood of the Variational Autoencoder (in Prose and Code).