Reconciliation of Mental Concepts with Graph Neural Networks

This repository accompanies the DEXA 2022 paper Reconciliation of Mental Concepts with Graph Neural Networks.

It contains all code as well as experimental setups described in the paper including results with al visualizations as standalone jupyter notebooks.

If you use code, data or any results in this repository, please cite:

@inproceedings{wendlinger2022reconciliation,
  title={Reconciliation of Mental Concepts with Graph Neural Networks},
  author={Wendlinger, Lorenz and H{\"u}bscher, Gerd and Ekelhart, Andreas and Granitzer, Michael},
  booktitle={International Conference on Database and Expert Systems Applications},
  pages={133--146},
  year={2022},
  organization={Springer}
}

Experiments

Complete experiments are stored in the notebooks for link prediction and basic network analysis.

Dataset

The TEAM-IP-1 Dataset described in the paper is also included in this repository.

Installation

Installation via the provided conda envirionment is encouraged.

conda env create -f abres_gcn.yml

To replicate the experiments, jupyter needs to be installed as well, e.g. with

conda install -c conda-forge notebook

or

pip install jupyterlab

Usage

All models and transformers are implemented as sklearn Estimators.

from link_prediction import LinkPredictor
import networkx as nx

# training graph
X_train: nx.DiGraph
# test graph to indicate potential edges
X_test : nx.DiGraph

abres_gcn = LinkPredictor()
abres_gcn.fit(X_train)
predictions = abres_gcn.predict(X_train, X_test)