Hi, On 08/30/2012 06:50 PM, cc wrote:
I know this is probably a very simple question to answer for those on here, but what are, or how does one determine the optimal values for the community_structure algorithm (http://projects.skewed.de/graph-tool/doc/community.html#graph_tool.community...
I know that the documentation states a history_file is available for this purpose, which I understand, but I'm confused as to what I am supposed to look for? Is there a certain pattern over a number of iterations that shows an optimal fit?
The algorithm attempts to minimize the Hamiltonian function defined in the documentation. Therefore the fit with the lowest value of this function should be considered the optimal fit.
Sorry if this is a bit confusing, but I'm very curious with respect to a kind-of tutorial on how to take the community_structure algorithm, decide the input values (on the basis of something?), and then determine which values fit the graph (again, on the basis of something else). My next step is, of course, to analyse Reichardt and Bornholdt theoretically to try and make sense of what I need to look for in this algorithm to apply it to something.
It looks like a very powerful algorithm, but the theoretical basics have me baffled. Thank you so much for your help!
The documentation is no substitute for knowing the literature... Community detection is a huge field, with new methods and approaches being developed almost in a weekly basis. I recommend reading the current review by Santo Fortunato: http://dx.doi.org/10.1016/j.physrep.2009.11.002 Cheers, Tiago -- Tiago de Paula Peixoto <tiago@skewed.de>