search

edwardlib / observations

Tools for loading standard data sets in machine learning
Other
202 stars 40 forks source link

Changed karate to use edgelist instead of gml #49

Closed zepx closed 6 years ago

zepx commented 6 years ago

The current karate.py uses read_gml which is not version agnostic. If networkx > 1.9.1 is used, then it breaks due to the changes in read_gml to support for labeling. Additionally, there's a bug with gml parsing in networkx from version 1.10 to 1.11.

Instead, I suggest using edgelist which is consistent across all networkx version. The dataset is obtained from another network repository: http://konect.uni-koblenz.de/networks/ucidata-zachary which has an edgelist version available instead of gml.

dustinvtran commented 6 years ago

Good catch! read_gml was a huge headache to work around. I'll merge this for now.

It would be great if you could also get this working for celegans.py. That's the only other networkx dependency at the moment. That way we can drop the version pre-req in setup.py.