An early attempt to combine sets and networks in a single visualization relied on first drawing an Euler diagram then placing a graph inside it [30], however the sets were often visualized with convoluted, difficult to follow curves. In addition, only limited kinds of set data could be shown as the system was limited to well-formed Euler diagrams. Compound graphs can be used to represent restricted kinds of grouped network data [8]. Graph clusters are visualized with transparent hulls by Santamaria and Theron [39]. However, the technique removes edges from the graph and it is not sufficiently sophisticated for arbitrary overlapping sets. Itoh et al. [24] proposed to overlay pie-like glyphs over the nodes in a graph to encode multiple categories. Each set is hence represented using disconnected regions that are linked by having the same colour. This causes difficulties with tasks that involve finding relations between sets such as T1, T3 and T4 in Section 5.3. A related class of techniques visualize grouping information over graphs using convex hulls, such as Vizster [22]. However, they do not support visualizing set overlaps.
