Networks, no matter how complex, are composed of simpler, smaller structures, called motifs. Some of these structures are easy to identify, like the pattern of a “star” where a single node acts as the sole connection to a connected component for one or more “pendant” nodes with a single tie. Another common pattern are nodes that are “parallel bridges” which share the only two connections they have with two or more other nodes. These common structures can be identified and removed and replaced with more efficient and comprehensible representations.
His paper with Prof. Ben Shneiderman at CHI 2013, “Motif simplification: improving network visualization readability with fan, connector, and clique glyphs“, demonstrates a novel method for improving the quality of network visualizations. Common network motifs appear frequently in networks. In network motif simplification these patterns are removed and replaced with simpler composite images:
The result is a simplification of the network visualization, removing clutter to reveal the core structural properties of interest.
This method for collapsing complex network graphs into simpler forms has been implemented in NodeXL. Look for the feature in the NodeXL Ribbon menu, in the NodeXL > Analysis > Groups > Group by Motif… option.
The feature allows users to select the types of motifs that should be recognized and collapsed:
Here is Dr. Dunne’s video explaining and demonstrating the concept:
For more information about the project, see: http://www.cs.umd.edu/hcil/nicernetvis