Social network spectra for benchmark test cases.

<p>Eigenvalues are all real because the networks are undirected. (a) Spectrum of Zachary Karate Club network data of 34 nodes from <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0054383#pone.0054383-Zachary1" target="_blank">[20]</a>. Inset (i) shows original adjacency matrix. Insets (ii) and (iii) show 2 dimensional and 4 dimensional approximations of original matrix. (b) Spectrum of Dolphins social network of 62 dolphins from <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0054383#pone.0054383-Lusseau1" target="_blank">[17]</a>. (c) Spectrum of American Football Teams social network from <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0054383#pone.0054383-Girvan1" target="_blank">[19]</a>. Left inset shows unordered original network data. Right inset shows the same network, where rows and columns have been reordered using algorithm in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0054383#pone.0054383-Sarkar1" target="_blank">[14]</a> to show modules along the main diagonal.