This paper is accepted to Information Systems [paper]🎉🎉

GenCAT allows users to generate attributed graphs with controlled relationships between classes, attributes, and topology.

GenCAT scales linearly to the number of edges and the number of attributes so massive attributed graphs can be generated.

The successor paper for an empirical study using GenCAT is accepted to NeurIPS 2022 Datasets and Benchmarks Track [paper] [code]. We highly recommend that you use the code for the NeurIPS paper.

• numpy == 1.20.2
• scipy == 1.6.2
• powerlaw == 1.4.6

(for demo)

• matplotlab == 3.4.1
• seaborn == 0.11.1

We show the demo in jupyter notebooks. You can make the same experiments to our paper by running the codes from top to bottom in the notebooks.