In another example, we consider the identification of the most central edges, where a central edge is defined as an edge traversed by the greatest number of shortest paths. One may compute all shortest paths using the algorithm of (Brandes 2008) and map its result to lines thickness in the node-link diagram.
In a last example, consider the detection of someone’s “influential social circles” (where the influence is let to be defined by the analyst) in a social network. A possible method consists in filtering (Ahlberg 1994) the network to highlight the nodes surrounding a selected node. But too many nodes are displayed if the node (or its direct neighbors) has a high number of connections. A solution is to define a function usually called “degree of interest” (Furnas 1986), which computes a score of how each node is related to the selected node, then to prune the visualization by keeping only the nodes of highest score. This method was used in (van Ham 2009) in another context.
Data mining algorithms may also be executed by interacting with the representation, like computing the shortest path after having selected the path endpoints. Integrating these algorithms into the visualization and making them available at any time of the exploration is thus a solution to include them in the non-linear processing chain.