In this repository, we present a continuous approximation of the Graph Edit Distance (GED) using the extended Sinkhorn algorithm. The evaluation utilizes three datasets: Acyclic, MAO, and RNA_trees. Within the 'ged.py' file, under the main section, users must specify the dataset to be used. By default, the Acyclic dataset is selected. To run the code, download the 'continuous_ged_approximation' folder and execute 'ged.py'. The implementation builds on the foundational work by Benoit Gaüzère and Luc Brun concerning the extended Sinkhorn algorithm. Components of their original code can be found in the 'deepged' folder. Additionally, parts of the Python package graphkit_learn, developed by Linlin Jia, Benoit Gaüzère, and Paul Honeine, have been incorporated and are available in the 'graphkit_learn' directory.

To use the uploaded Jupyter Notebook 'ged_analysis' for plotting: download continuous_ged.csv from https://drive.google.com/file/d/1NraU6g7NWPFq1557DhCqZcMzFXfOxtvP/view?usp=share_link and save it in the folder create_plots.