Home

This repository contains code to process weighted graphs from Netzschleuder to export them in a CSV format supported by the ERODE tool

Please open the following jupyter lab notebook on Google colab and run it.

• Please click on this badge:

After following the link:

• Click on 'Copy to Drive' to create your own copy of this file on your Google drive

  • highly suggested, otherwise you will loose any change applied and file created
  • If you decide to copy the notebook, the next time open your local copy by accessing it from here, navigating to Colab Notebooks

• Add all graphs of interest in the list model_names

  • model_names=["blumenau_drug"]
  • In most of the cases, you shall use the identifiers in the first column here
    • For some models you might need to use a different ID, as specified in the repository page
    • E.g., ambassador actually consist of 15 graphs, each of which shall be named as ambassador/model_name1, ambassador/model_name2 ...
  • In the notebook you find a list of all weighted graphs found on 21/07/22 sorted from smallest number of nodes to largest
  • E.g.:
    • model_names=["blumenau_drug","cattle"]

• Set flags download_file and store_on_gdrive to True or False

  • store_on_gdrive=True implies that exported CSV files will be stored in your google drive
  • store_on_gdrive=True (and at the same time store_on_gdrive=False) implies that exported CSV files will be downloaded

• Run all cells.

  • All installation steps will be performed
  • The weighted graphs specified in list model_names will be processed
    • Colab might fail if you try to process all weighted graphs due to excessive usage of RAM (just split the list in two or more, and run again)
      • Certain models are just too big for colab (e.g. bag_of_words/pubmed)

Output

• You will get in output:
  • model_nameA.csv: a compact representation of the edges: source,target,weight
  • model_nameB.csv: a file necessary to ERODE. It contains value 0 for each node
  • import_model_name.ode: an ERODE file able to process the CSV files. It will create a file containing the ODEs corresponding to the system of equations Ax=B.
• You shall:
  • Download and run ERODE
  • Create a new project
  • Drag&drop the three files there. Open import_model_name.ode, and click on the button with the ERODE logo to run the ERODE file
  • Please find manual and installation instructions here

Beware: not all graphs classified as weighted actually come with weights.

• E.g., hens does not actually come with weights.
• In these cases, we
  • set all weights to 1
  • add 'defaultWeights' to the name of every created file
• Other models have custom names for the 'weights' column in the source files
  • In the jupyter file you will find this handled for some classes of models (dictionary prefix_to_weight)