This repository is associated with the paper "Neuro-Symbolic Learning of Answer Set Programs from Raw Data", published at IJCAI 2023. A recorded presentation is available. Please consider citing if it is useful in your work. The technical appendix forming our supplementary material is available here.

If running on an intel machine, the recommended approach is using Docker. For Apple Silicon users, Docker can be used by adding the --platform linux/amd64 flag to the build and run commands, however, the performance hit is significant. It is recommended to install natively.

A docker container is provided for easy setup and installation so the only additional software required is a container runtime such as Docker Desktop, Podman, or minikube.

The following commands assume the docker command is available on your system. If using a different container run-time, please replace accordingly. From the root directory:

1. docker build -t nsil:ijcai_2023 .
2. docker run -d -p 8000:8000 -p 9990:9990 --name nsil_ijcai_2023 nsil:ijcai_2023
3. Open http://localhost:8000 in your web browser to launch a markdown viewer containing detailed documentation that explains how to reproduce the experiment results.

Once the docker container is running:

1. docker exec -it nsil_ijcai_2023 /bin/bash
2. cd scripts
3. Run experiments using the commands detailed in the documentation.

exit
docker stop nsil_ijcai_2023
docker rm nsil_ijcai_2023
docker rmi nsil:ijcai_2023

• Python >= 3.9 with Pip
• Clingo with lua support
• (Optional) virtualenv or conda.

1. Create python environment and install dependencies

virtualenv nsil_p3
source nsil_p3/bin/activate
pip install -r requirements.txt

2. Download image data

chmod +x ./download_data.sh
./download_data.sh

3. Create a setup script with the following contents:

BASE_PATH=/path/to/NSIL
export PYTHONPATH=$BASE_PATH
cd $BASE_PATH/examples/$1

and replace the value of BASE_PATH accordingly. Save this file to paper_experiments/ijcai_2023/scripts/setup.sh and ensure it has execute permissions. If you also want to run the NeurASP and FF-NSL baseline experiments, save a copy to paper_experiments/ijcai_2023/scripts/naive_baselines/setup.sh, again ensuring executable permissions.

4. Move ILASP and FastLAS binaries to the root directory. We provide apple silicon and intel ubuntu linux binaries, copy accordingly.

mv LAS_binaries/ILASP_apple_silicon ILASP
mv LAS_binaries/FastLAS_apple_silicon FastLAS

5. Set the python path to the root directory

export PYTHONPATH=/path/to/NSIL

6. To view documentation:

cd paper_experiments/ijcai_2023/
chmod +x ./start_web_servers.sh
./start_web_servers.sh

7. To run experiments:

cd paper_experiments/ijcai_2023/scripts
find . -type f -exec chmod +x {} \;
./run_arithmetic_repeats.sh -p 100 -s 0 -m 5 # For example - see documentation for more commands.

TBC