sbm-bp implements the belief propagation algorithm for the inference of the (degree-corrected) stochastic block model. This program is largely an object-oriented re-implemetation of the "MODE-NET" code by Aurelien Decelle, Florent Krzakala and Pan Zhang.
This program is tested with a few benchmark networks, including synthetic networks and the karate club network.
It generates nearly identical outputs as the original ones. For completeness, the original code is also included in this repository, under src/old.
Documentation will be updated soon.
This code requires compilers that support C++14 features.
It also depends on boost::program_options and cmake.
Compilation:
cmake .
make
The binaries are built in bin/.
bin/bp -l dataset/N_1000-Q_2-method_cab_ec-eps_0.1-c_3.0.edgelist -n 500 500 --pa 0.5 0.5 --cab 3.63 2.36 3.63 -t 1000 -i 0 --deg_corr_flag 0 -m infer
bin/bp -l dataset/N_1000-Q_2-method_cab_ec-eps_0.1-c_3.0.edgelist -n 500 500 --pa 0.5 0.5 --cab 3.63 2.36 3.63 -t 1000 -i 0 --deg_corr_flag 0 -m learn
Please find the references from the original project page.