Interactive Visualization of Huge Graphs Using Kernel Functions for Aggregation

Abstract

I propose a new graph visualization technique that allows the interactive exploration of large datasets with adjustable level of detail. The approach is designed to exploit the architecture of common graphics hardware and achieves interactive speed through parallelization. To address information hiding through node and edge clutter, I replace the standard node link diagram with a novel combination of a continous node-density visualization with discrete meta-edges. Additionally, I present a zoom & pan based user interface that allows the intuitive exploration of the data at distinct aggregation levels and supports filter and detail on demand mechanisms to address the characteristics of different datasets and research questions.