Grooming the hairball - how to tidy up network visualizations?

摘要: Every visualization researcher and practitioner knows the painful experience of a beautifully designed network layout breaking down once input graph scales up to realistic node edge counts. The resulting "hairball" suffers from cluttering overplotting an extreme that renders it unusable for any practical purposes. Since researchers have had this decades, various approaches been developed on all stages pipeline alleviate problem. They range filtering clustering techniques data level modern GPU-based image level. This tutorial gives overview these discusses their applicability interplay in different application scenarios. By doing so, provides unique problem-oriented perspective field scalable visualization, which is area active research today more than ever. serves mainly further understanding beyond point creating initial layout. It thus caters intermediate audience with some basic knowledge but will certainly present interesting cross-section through larger domains drawing established as well.