This package contains a set of methods to calculate:

1. Renormalizable Vector Embeddings, if the full adjacency matrix is known. See "Multi-Scale Node Embeddings for Graph Modeling and Generation" arXiv;

2. Global or local scalar embeddings, if only marginal quantities are available, such as the number of links or the node degrees. See "Renormalizable graph embeddings for multi-scale network reconstruction" arxiv

Feel free to reach out at riccardo.milocco@imtlucca.it or riccardo.milocco@gmail.com, for code refinement, feedback, or collaborations.

Note: to extend the methods to the directed case, rely on the same pipeline, while doubling the number of vectors: instead of having only one $\vec{x}_i$ per node, introduce $\vec{x}_i, \vec{y}_i$ that will be interpreted as the ``propensity'' of sending a message/interaction and receiving it resp (see here). This is largely different from taking 2*dimX. Thus, it needs to be carefully implemented by following the previous steps.

Install from build:

python3 -m venv .venv pip3 install --editable . --config-settings editable_mode=compat 

Please work on a feature branch and create a pull request to the development branch. If necessary to merge manually do so without fast forward:

git merge --no-ff myfeature 

To build a development environment run (Linux):

python3 -m venv .venv source .venv/bin/activate pip3 install --editable . --config-settings editable_mode=compat 

Have fun! 🚀