Code in support of "Direct Evolutionary Optimization of Variational Autoencoders With Binary Latents"

Overview

The experiments directory contains implementations of the experiments described in the paper. Execution requires an installation of the Truncated Variational Optimization (TVO) framework, which implements the Truncated Variational Autoencoder. Experiments furthermore leverage pre-/postprocessing and visualization utilities provided by tvutil.

After following the Setup instructions described below, you will be able to turn to running the experiments. Please consult the READMEs in the experiments' sub-directories for further instructions.

The code has only been tested on Linux systems.

Setup

We recommend Anaconda to manage the installation, and to create a new environment for hosting the installed packages:

$ conda env create
$ conda activate ecml2022

The tvo package can be installed via:

$ git clone https://github.com/tvlearn/tvo.git
$ cd tvo
$ python setup.py build_ext
$ python setup.py install
$ cd ..

To install tvutil, run:

$ git clone https://github.com/tvlearn/tvutil.git
$ cd tvutil
$ python setup.py install
$ cd ..

Reference

@InProceedings{DrefsGuiraudEtAl2022,
  author="Drefs, Jakob
  and Guiraud, Enrico
  and Panagiotou, Filippos
  and L{\"u}cke, J{\"o}rg",
  editor="Amini, Massih-Reza
  and Canu, St{\'e}phane
  and Fischer, Asja
  and Guns, Tias
  and Kralj Novak, Petra
  and Tsoumakas, Grigorios",
  title="Direct Evolutionary Optimization of Variational Autoencoders with Binary Latents",
  booktitle="Machine Learning and Knowledge Discovery in Databases",
  year="2023",
  publisher="Springer Nature Switzerland",
  address="Cham",
  pages="357--372",
}