Improving resolution of dynamic communities in human brain networks through targeted node removal.

ABSTRACT: Current approaches to dynamic community detection in complex networks can fail to identify multi-scale community structure, or to resolve key features of community dynamics. We propose a targeted node removal technique to improve the resolution of community detection. Using synthetic oscillator networks with well-defined "ground truth" communities, we quantify the community detection performance of a common modularity maximization algorithm. We show that the performance of the algorithm on communities of a given size deteriorates when these communities are embedded in multi-scale networks with communities of different sizes, compared to the performance in a single-scale network. We demonstrate that targeted node removal during community detection improves performance on multi-scale network

SUBMITTER: Schlesinger KJ 

PROVIDER: S-EPMC5737970 | biostudies-literature | 2017

REPOSITORIES: biostudies-literature