graspe

Graphs in Space: Graph Embeddings for Machine Learning on Complex Data

This repository contains code for Graph Embedding evaluation for the GRASP project.

Structure

graspe
├── conda-env.yml       -> conda package manager environment definition
├── data                -> dataset files
├── LICENSE             -> project license file
├── models              -> saved models that can be loaded from disk
├── notebooks           -> jupyter notebooks
├── README.md           -> project readme
├── reports             -> documentation, instructions, manuals, reports, plots, etc.
├── requirements.txt    -> pip package manager environment definition
└── src/{graspe,tests}  -> source code files, and test files

Installation

We recommend using this library by installing conda via conda-env.yml file. You can do that in two steps:

• a. Conda: conda env create -f conda-env.yml && conda activate graspe
• b. pip install -r requirements.txt (In clear environment)

Examples

Running tests

cd src/graspe && pytest -rP

Generate pydoc

cd src/graspe
pydoc -w `find . -name '*.py'`
mv *.html ../../reports/doc

Graph Auto Encoders

python src/graspe embed gae -g karate_club_graph -d 10 
python src/graspe embed gae -g karate_club_graph -d 10 --variational # to use VAE 

Example of graph embedding file

out.embedding

Authors

(c) 2020 UNSPMF

License and Acknowledgements

• GNU General Public License v3.0
• This research (library) is supported by the Science Fund of the Republic of Serbia, #6518241, AI -- GRASP.