Graph Exploration With Embedding-Guided Layouts

Area of Science:

• Computer Science
• Data Visualization
• Graph Theory

Background:

• Node-link diagrams are essential for graph visualization, but existing layout algorithms struggle to balance aesthetic goals (e.g., minimizing edge crossings) with exploration goals (e.g., preserving community structures).
• Current hybrid methods often face limitations such as restricted input types, manual adjustments, and a lack of balance between visual appeal and analytical insight.

Purpose of the Study:

• To propose a flexible embedding-based graph exploration pipeline that effectively integrates graph topology and node attributes.
• To develop an embedding-driven graph layout algorithm (GEGraph) that enhances both aesthetic quality and community preservation for improved graph interpretation.
• To extend graph exploration capabilities through layout-preserving aggregation and proximity-based node searching.

Main Methods:

• Leveraged embedding algorithms for attributed graphs to encode both graph topology and node attributes into a latent space.
• Developed GEGraph, an embedding-driven graph layout algorithm designed for aesthetic appeal and superior community preservation.
• Implemented layout-preserving aggregation with Focus+Context interaction and a related nodes searching approach utilizing multiple proximity strategies.

Main Results:

• The proposed pipeline successfully integrates graph topology and node attributes, overcoming limitations of existing methods.
• GEGraph generates aesthetically pleasing graph layouts that effectively preserve community structures, facilitating easier interpretation.
• The developed exploration methods enhance usability and provide deeper insights into graph data.

Conclusions:

• The embedding-based graph exploration pipeline offers a flexible and effective approach to visualizing and analyzing attributed graphs.
• GEGraph and associated exploration tools provide a significant advancement in balancing aesthetic layout goals with analytical requirements for graph data.
• Evaluations, including a user study and case studies, validate the approach's effectiveness and practical utility.