Multicriteria Scalable Graph Drawing via Stochastic Gradient Descent, (SGD)²

Area of Science:

• Computer Science
• Graph Theory
• Data Visualization

Background:

• Graph readability is crucial for data comprehension.
• Existing algorithms often optimize single criteria, sacrificing others.

Purpose of the Study:

• Introduce a novel approach for optimizing multiple graph readability criteria.
• Enable flexible and scalable graph layout generation.

Main Methods:

• Propose Multicriteria Scalable Graph Drawing via Stochastic Gradient Descent (SGD^2).
• Utilize differentiable functions to optimize diverse criteria like edge length, stress, and neighborhood preservation.
• Adapt the method for planar graph optimization while maintaining planarity.

Main Results:

• (SGD)^2 handles multiple criteria, including novel ones like node and angular resolution.
• The approach demonstrates scalability for large graphs.
• Quantitative and qualitative analyses validate layout quality, runtime, and criterion interactions.

Conclusions:

• (SGD)^2 offers a flexible, scalable solution for complex graph layout optimization.
• The method advances graph visualization by integrating multiple readability objectives.
• Open-source code and interactive demos facilitate adoption and further research.