UnlearnSPN

June 7, 2022 ยท View on GitHub

Installation

  1. Create a new conda environment from the conda.yml file:
conda env create -n unlearnspn --file conda.yml
  1. Activate the conda environment:
conda activate unlearnspn
  1. Unpack the modified version of SPFlow:
unzip SPFlow.zip
  1. Navigate to SPFlow source directory and install it manually:
cd SPFlow/src/
python setup.py install
  1. Unzip datasets:
unzip data.zip

Reproducing the results

In order to reproduce the results from the paper, simply run:

./run_training_comparison.sh && ./run_experiments.sh

The results will be stored using mlflow. In order to view the results run

mlflow ui

in the directory containing the mlruns directory. This will start a local web server, which can be accessed through 127.0.0.1:5000 in your web browser.