mbisbm
stands for Matched BIpartite Stochastic Block Model.
The code here implements algorithms in the paper entitled
Matched bipartite block model with covariates.
The main function is common/fit_mbiSBM.m
which implements the latest version of the algorithm.
This is the defulat version and should be used if you have a good initialization.
See example_wiki_cities.m
and example_wiki_topart.m
for how to use this function.
common/biSpecClust.m
implements the matched bipartite clustering described in the paper.
This can be used as the initialization for the mbisbm or as a standalone algorithm.
There is also common/fit_mbiSBM_v2.m
which implements v.2 of the algorithm. This is used for example
in creating some of the early figures in the paper. You can run simulations/compare_nmi_sims.m
to see it in action. V.2 is more robust to the initial labeling.
Use this version if you have unreliable/bad initial labels.