Community detection ... in pictures
7 - 7
Community detection identifies well-connected groups of people (or anything) based on the known relationships between them.
As simple as the concept is, some community detection solutions can have problems discerning small groups with large data sets.
Our approach is one that circumvents this problem.
The multi-scale algorithm can reliably identify hierarchical levels in networks such as this one.
We developed a multi-scale algorithm that compares communities against separately-solved copies of the system.
This enabled us to reliably and quantitatively evaluate the "best" resolutions for most networks without a priori knowledge of the network and without forcing an anticipated structure .
Our multi-scale algorithm compares different solutions of the system, looking for how well they agree or disagree in order to evaluate the solutions.
Among other applications, we applied our methods to find clusters in amorphous solids.
The multi-scale algorithm identifies relevant structural scales for lattices, but ironically, symmetric systems such as this cube are often more difficult to evaluate than typical community detection problems.
Unless otherwise stated, all content and images are copyright © 2013 Peter Ronhovde
Images of the Sun are from NASA archives.