Skip to content

Latest commit

 

History

History
107 lines (83 loc) · 4.86 KB

README.md

File metadata and controls

107 lines (83 loc) · 4.86 KB

############################################################################### Copyright (C) 2012, M. Schaub, M. Barahona

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.

###############################################################################


Community Detection using Markov Zooming Infomap

This code implements a MATLAB/C++ version of the Markov Zooming Map equation method as discussed in the article:

"Encoding dynamics for multiscale community detection: Markov time sweeping for the Map equation", M.T.Schaub, R. Lambiotte, M. Barahona, Phys. Rev. E, 2012, 86(2), 026112; see also arXiv:1109.6642

The code implements the Markov Zooming Map equation for undirected, weighted networks and is based on the implementation of the original map equation method which can be found under http://www.tp.umu.se/~rosvall/code.html See also http://www.mapequation.org for a dynamic visualisation of the map equation. The original Map equation method is described in the following publication:

"Maps of information flow reveal community structure in complex networks" Martin Rosvall and Carl T. Bergstrom, PNAS 105, 1118 (2008); see also arXiv:0707.0609

Please cite the above mentioned articles if you make use of this code.


INSTALLATION & USE

  1. Download the files and unzip the folders it neccessary
  2. in a terminal, go into the infomap_dir folder and run "make" to compile the C code.
  3. In Matlab you can now use the function "MarkovZoomingMap" to run your analysis

EXAMPLE:

 % in the MarkovZoomingMap folder, load the provided example graph 
 % (a ring of rings) into an adjacency matrix
 A = convertPajekToAdjMatrix('ring_of_rings.net');
 % COMMENT: If want to read in an unweighted pajek graph 
 % (edges encoded: [from] [to] -- with no third colum)
 % use the function "convertUnweightedPajekToAdjMatrix".
 
 % assign an output filename
 filename = 'test';

 % specify a time interval for the analysis
 time =logspace(-1,2,100);

 % run the actual analysis; note that the time argument is optional, if not
 % provided the time interval is set to the one also used in this example
 MarkovZoomingMap(A,filename,time)

This should create a folder named "{filename}ZoomingMap" containing the results of the analysis in a file named Map_clustering.mat, plus a (directed) pajek .net file of the original graph.

To plot the results of the analysis you may want to load the .mat file and use the provided matlab script "script_plot_Map_results.m".


Some potential code changes / improvements

This released code has not been thoroughly optimized for speed, however there are two obvious ways to speed up the performance.

  1. At the moment for each timepoint, a temporary pajek file is created and written to disk, and subseqently read from the C code. This I/O is essentially not necessary. Instead on could pass the network data straight to the C implementation via e.g. a mex file from Matlab (However if you are interested in the "intermediate" graphs you may not want to do this, and instead change the script as to not delete the temporary files but keep them).

Further you may want to stop all the output from being displayed while the optimization is running, this is also easily achievable by changing the C code accordingly.


Code history

Aug. 2012 -- intial release of code
Nov. 2013 -- updated code to compute stationary distribution quicker, included option to use linearisation of dynamics (equivalent to original map equation for t=1)


Author : M. Schaub
Email : michael.schaub09@imperial.ac.uk