Interactive visualization of large graphs and networks

摘要: Many real-world domains can be represented as large node-link graphs: backbone Internet routers connect with 70,000 other hosts, mid-sized Web servers handle between 20,000 and 200,000 hyperlinked documents, dictionaries contain millions of words defined in terms each other. Computational manipulation such graphs is common, but previous tools for graph visualization have been limited to datasets a few thousand nodes. Visual depictions networks are external representations that exploit human visual processing reduce the cognitive load many tasks require understanding global or local structure. We assert two key advantages computer-based systems information over traditional paper-based exposition interactivity scalability. also argue designing software by taking characteristics target user's task domain into account leads more effective scale larger than work. This thesis contains detailed analysis three specialized interactive exploration graphs, relating intended spatial layout encoding choices. present novel algorithms drawing use quite different metaphors. The H3 system visualizing hyperlink structures web sites scales 100,000 nodes using carefully chosen spanning tree backbone, 3D hyperbolic geometry Focus+Context view, provides fluid experience through guaranteed frame rate drawing. Constellation features highly 2D spatially encode domain-specific computational linguists checking plausibility semantic network created from dictionaries. Planet Multicast displaying tunnel topology Internet's multicast literal geographic arcs on globe help MBone maintainers find misconfigured long-distance tunnels. Each these very view structure, we evaluate their efficacy task. generalize findings our importance specialization scalable.