GraphMaps: Browsing Large Graphs as Interactive Maps

Abstract

Algorithms for laying out large graphs have seen significant progress in the past decade. However, browsing large graphs remains a challenge. Rendering thousands of graphical elements at once often results in a cluttered image, and navigating these elements naively can cause disorientation. To address this challenge we propose a method called GraphMaps, mimicking the browsing experience of online geographic maps. GraphMaps creates a sequence of layers, where each layer refines the previous one. During graph browsing, GraphMaps chooses the layer corresponding to the zoom level, and renders only those entities of the layer that intersect the current viewport. The result is that, regardless of the graph size, the number of entities rendered at each view does not exceed a predefined threshold, yet all graph elements can be explored by the standard zoom and pan operations. GraphMaps preprocesses a graph in such a way that during browsing, the geometry of the entities is stable, and the viewer is responsive. Our case studies indicate that GraphMaps is useful in gaining an overview of a large graph, and also in exploring a graph on a finer level of detail.

Simplified Emanation Graphs: A Sparse Plane Spanner with Steiner Points

“…A theoretical open question is to prove a tight upper bound on spanning ratio of the simplified version. Furthermore, one can implement simplified emanation graphs in visualization systems such as GraphMaps [13] to compare the visual results with that of generated by the Delaunay and constrained Delaunay triangulations.…”

“…Many applications use t-spanners, and in general, planar geometric graphs, in different applied areas of computational geometry and data visualization. Nachmanson et al [13] introduced a system called GraphMaps for interactive visualization of large graphs based on constrained Delaunay triangulations. Later, Mondal and Nachmanson [12] introduced and used a specific mesh called the competition mesh to improve GraphMaps (Figure 1).…”

A New Approach to GraphMaps, a System Browsing Large Graphs as Interactive Maps

“…The GraphMaps system proposed previously (Nachmanson et al, 2015) uses 1 to obtain the graph for routing edges on a level. Our approach does not depend on Triangle, but uses Competition Mesh.…”

“…Other forms of clutter reduction approaches include node aggregation (Wattenberg, 2006;Dunne and Shneiderman, 2013;Zinsmaier et al, 2012), topology compression (Shi et al, 2013;Brunel et al, 2014), and sampling algorithms (Gao et al, 2014). This paper focuses on GraphMaps, proposed by Nachmanson et al (Nachmanson et al, 2015), that reduces clutter by distributing nodes to different zoom levels and routing edges on shared rails. Like the clutter reduction approaches, a primary goal of GraphMaps is to make the visualization more readable and interactive in the higher levels of abstraction.…”

“…We build the tiles as in (Nachmanson et al, 2015). In the first level we have only one tile coinciding with the graph bounding box.…”