Matrix BrowserVisualizing Networked Informationpresented bySandro FouchéTypical Solution (nets as nets)Problem‣Overview and Search in large networks is difficult‣Visual Search is impeded by arbitrary positioning of nodes‣Associated nodes are often placed far from neighbors‣Finding larger patterns or following long paths becomes difficultState of the Art(2001)Hyperbolic Trees‣Better for hierarchical data‣User loses context because focus shifts‣Hard to follow a specific path‣Limited search and organizing toolsMatrix-view of Networks (Becker ‘95)Matrix Browser‣Grouped hierarchically‣Organized using a standard tree-widget ‣Sortable‣(Limited) filteringNodes‣Represented by intersections between nodes on each axis called relations‣Higher level relations are inferred to make data hiding feasible (explicit, hidden, implicit inheritance / generalization)‣Can be uni-directional, bi-directional, or directionlessArcsDemoBenefits‣Uses known metaphors‣High interactivity‣User-controlled filtering‣Structured organization of dataResults‣vs. unstructured network graphs:50% search time (performance times)50% visual fixations (eye-tracking)‣vs. strong visual structured graphs:15% performance advantage (w/o interactive features)ConclusionsPros‣manages large number of nodes and arcs well‣Easy to find a given node or relationship‣Collapses detail while maintaining structural informationConclusionsCons‣Doesn’t handle spatial locality as well as one would like‣Tool seems “not ready for prime-time”Q &
View Full Document
Unlocking...