The map generator

We have developed the Flash applet above to make it easier for you to simplify and highlight important structures in your networks with the map equation (we recommend that you use the latest version of FlashPlayer). When you load your weighted or unweighted, directed or undirected network in Pajek format into the applet, the map generator clusters the network based on the map equation and generates a map for you to customize and save as a pdf file. A visualization of the mechanics of the map equation is available here, a detailed description is available here and C++ source code to partition large networks with millions of weighted directed links is available here.

References

M. Rosvall and C. Bergstrom, Maps of information flow reveal community structure in complex networks,
PNAS 105, 1118 (2008). [pdf] [arXiv:0707.0609]

Martin Rosvall, Daniel Axelsson, and Carl T. Bergstrom, The map equation,
Eur. Phys. J. Special Topics 178, 13 (2009). [pdf] [arXiv:0906.1405]

File formats

.net
Structure your network into Pajek's .net format. Here is an example network with three nodes and four directed and weighted links:

		 *Vertices 6
		 1 "Node 1" 1.0
		 2 "Node 2" 1.0 
		 3 "Node 3" 1.0 
		 4 "Node 4" 1.0 
		 5 "Node 5" 1.0 
		 6 "Node 6" 1.0
		 *Arcs 8
		 1 2 2.0
		 2 3 2.0
		 3 1 2.0
		 1 4 1.0
		 4 5 2.0
		 5 6 2.0
		 6 4 2.0
		 4 1 1.0	

Pajek uses "Edges" for undirected links and "Arcs" for directed links. We accept both "Edges" and "Arcs" and the choice of load button determines whether the algorithm treats the network as undirected or directed. Directed links have the form "from to weight". That is, the first link in the list above goes from node 1 to node 2 and has weight 1.9. The link weight is optional and the default value is 1 (we aggregate the weights of links defined more than once). Node weights are optional and sets the relative proportion to which each node receives teleporting random walkers in the directed version of the code. Default value is 1.

.clu
Structure your clustering file in Pajek's .clu format, which is just a list of module assignments:

		 *Vertices 6
		 2
		 2
		 2
		 1
		 1
		 1			

.map
For very large networks, the load and clustering time can be long in the Flash-based Map Generator. To overcome this problem, the clustering algorithm can be run offline with the code provided here. The code outputs a file with extension .map:

		# modules: 2
		# modulelinks: 2
		# nodes: 6
		# links: 8
		# codelength: 2.51912
		*Directed
		*Modules 2
		1 "Node 1" 0.5 0.0697722
		2 "Node 4" 0.5 0.0697722
		*Nodes 6
		1:1 "Node 1" 0.209317
		1:2 "Node 3" 0.147071
		1:3 "Node 2" 0.143613
		2:1 "Node 4" 0.209317
		2:2 "Node 6" 0.147071
		2:3 "Node 5" 0.143613
		*Links 2
		1 2 0.0697722
		2 1 0.0697722

This file only contains the necessary information to generate a map. The names under *Modules are derived from the node with the highest PageRank within the module and 0.5 and 0.0697722 represent, respectively, the aggregated PageRank of all nodes within the module and the per step exit flow from the module. The nodes are listed with their module assignments together with their PageRanks. Finally, all links between the modules are listed in order from high flow to low flow.

Instructions

1. Load the .net network file (see left column for details) into the map generator by clicking on the button "Load network" and chose between undirected and directed network. For very large networks, the load and clustering time can be long in the Flash-based Map Generator. If you are encountering problems, you can run the clustering code offline and load the .map file into the applet (see .map under File formats in the left column).
   
   If you just want to try out the map generator, we have provided a few sample networks. The weighted undirected network is the largest connected component of a coauthorship network compiled (for details of how weights are assigned, see M. E. J. Newman, Phys. Rev. E 64, 016132 (2001)) from two network reviews and one community detection review: M. E. J. Newman, SIAM Review 45, 167-256 (2003); S. Boccaletti et al., Physics Reports 424, 175-308 (2006); S. Fortunato, Physics Reports 486, 75-174 (2010), and can be downloaded in .net format here. The networks of science come from Thomson-Reuters’ Journal Citation Reports 2004. Our data tally on a journal-by-journal basis the citations from articles published in a given year to articles published in the previous five years, and self-citations are excluded. See Eigenfactor.org for more information about ranking of scientific knowledge.
   
   
2. Cluster the network based on the map equation or alternatively provide a clustering in Pajek's .clu format. The algorithm tries to minimize the description length of a random walker's movement on the network and reveals important aspects of network structure with respect to the dynamics on the network. (More information here.)
   
   
3. The map generator displays the network as a map. Every module represents a cluster of nodes and the links between the modules represent the flow between the modules. The size of a module is proportional to the average time a random walker spends on nodes in the module and the width of a link is proportional to the per step probability that a random walker moves between the modules:
   
   
4. Customize the map by changing the position of the modules automatically (see "Placement tools" in the navigator) or manually, by changing the names of the modules (double click on a module to see the nodes within the module), by changing the color of modules and links (see "Color and size tools" in the navigator), by moving the labels, etc. All adjustments can also be applied only to selected modules (shift-click selects a single module and shift-drag selcts multiple modules).
   
   
5. Save the customized map in scalable vector graphics as a pdf file or as a .map file for later access in the Map Generator (see "Network input/output" in the navigator).