Welcome to Geo2DR’s Docs!

Geo2DR is a Python library for constructing methods capable of learning distributed representations of graphs. Here, embeddings of substructure patterns (walks, graphlets, trees, etc.) and whole graphs are learned by exploiting the distributive hypothesis used in statistical language modelling. It attempts to make the various algorithms used for inducing discrete structures, creating corpi, training neural language models available under a simple easy to use library. This will allow the rapid recreation of existing methods as well as construction of completely novel unpublished methods in a robust and reliable manner.


This is an actively developing project and a lot more documentation is planned to be included (you can view the raw rst file to see the upcoming tutorials and reference materials to be included in the coming weeks). It’s quite tough going handling this project alone, so please bear with me!

If the code or specific examples of existing systems is useful to your work, please cite the original works and also consider citing Geo2DR.

      title={Learning distributed representations of graphs with Geo2DR},
      author={Paul Scherer and Pietro Lio},