Node Embeddings Survey

Authors

This is a survey on Node Embeddings from Christos Ziakas, Jan Rüttinger, and Till Richter. It was conducted in the Machine Learning Lab of the Data Analytics and Machine Learning Group from TUM. We thank Oleksandr Shchur for supervising our project and Prof. Dr. Günnemann for the possibility to conduct research at his group.

What is the project about?

This project compares different node embeddings with one shared evaluation protocol.

Architecture of the project

/embedding

• contains the implementation of 3 embedding methods
• contains the implementation of several embedding methods

/evaluation

• contains the implementation of 3 evaluation tasks

/experiment

• contains 3 jupyter notebooks two run and visualize experiments

/gust

• contains a helper library developed by the chair to load and preprocess data

/utils

• contains helper code

How to run the code?

1. Install all requirements
   pip install -r requirements.txt

2. Define and run an experiment
   Open the jupyter notebook Experiment_pipeline.ipynb in the folder "experiments" and follow its instructions.

3. Visualize the results
   Open the jupyter notebook Visualize_Results.ipynb in the folder "experiments" and follow its instructions.