This repo contains the files that implemented the experiments and the data used in the paper 'Deep Learning for Link Prediction in Realistic Biomedical Graphs: A Multidimensional Evaluation of Graph Embedding-based Approaches' by Gamal Crichton, Yufan Guo, Sampo Pyysalo and Anna Korhonen.
The experiment_batch.sh file contains the steps to automate the experiments. The .py files are used in the batch script. The folders LINE, DeepWalk, node2vec etc. contains implementations of those algorithms obtained freely on the internet. The data folder contains the files for the graphs used. In cases where the files are split into numbered segments, combine them before doing the experiments. They were only split to get around Github's file size limit. The files are generally well-documented and the parameters that each script accepts are in the files.
Feel free to open an issue if anything does not work.
Dependencies
- Python 2.7
- Tensorflow and tflearn
- Networkx
- SciKit-learn
- GEM
- Numpy
- Pandas
The code is provided under MIT license and the other materials under Creative Commons Attribution 4.0.