VAEL

Codebase for VAEL: Bridging Variational Autoencoders and Probabilistic Logic Programming published at NeurIPS 2022.

If you use this codebase, please cite:

@inproceedings{NEURIPS2022_1e38b2a0,
 author = {Misino, Eleonora and Marra, Giuseppe and Sansone, Emanuele},
 booktitle = {Advances in Neural Information Processing Systems},
 editor = {S. Koyejo and S. Mohamed and A. Agarwal and D. Belgrave and K. Cho and A. Oh},
 pages = {4667--4679},
 publisher = {Curran Associates, Inc.},
 title = {VAEL: Bridging Variational Autoencoders and Probabilistic Logic Programming},
 url = {https://proceedings.neurips.cc/paper_files/paper/2022/file/1e38b2a0b77541b14a3315c99697b835-Paper-Conference.pdf},
 volume = {35},
 year = {2022}
}

Prerequisites

• Python >=3.7
• Dependencies:

  pip install -r requirements.txt

  Note: if something goes wrong with PySDD, try pip install -vvv --upgrade --force-reinstall --no-binary :all: --no-deps pysdd

Usage

1. Clone the repo

   git clone https://github.com/EleMisi/VAEL.git

2. Install the dependencies

   pip install -r requirements.txt

3. Set the experiment(s) configuration in file config.py

4. Run the experiment(s)

   python run_VAEL.py

   Use flag --task mnist to run 2digit MNIST experiment(s), and --task mario to run Mario experiment(s).

Results

The results are stored in the folder ./<exp_folder>/<exp_class>/ specified in run_VAEL.py.

In particular:

• the resulting metrics for each tested configuration are reported in exp_class.csv
• each subfolder refers to a specific configuration and contains
  • the model checkpoint
  • the learning curves
  • some samples of image reconstruction and generation

Corresponding Author

• Eleonora Misino - eleonora.misino2@unibo.it
• Giuseppe Marra - giuseppe.marra@kuleuven.be
• Emanuele Sansone - emanuele.sansone@kuleuven.be