zhuloucao/collective_influence

Collective Influence

Implementation of the collective influence algorithm as presented by Morone and Makse (2015). The collective influence computation step has been parallelized for use with large networks.

Depends on igraph.

Quick sourcing of the script

If you want to test this out quickly:

source("collective_influence_algorithm.R")
plotg(exampleGraph(), community=T)
# find influencers, remove them, and plot the results after each removal
g <- getInfluencers(exampleGraph(), d=2, plot=T)

Note that getInfluencers() is parallelized , which can improve processing time for collective influence on large networks when multiple processing cores are available. Simply specify the number of cores to use.

More information

Example from Kovacs and Barabasi News & Views article (2015): (A) the initial network with highest collective influence node in red, highest degree node in yellow. (B) resultant network, with giant component intact, after removal of the 6 nodes with the highest degree. (C) resultant network after removal of top 4 influencers.

The CI algorithm at work: