What Would a Graph Look Like in this Layout? A Machine Learning Approach to Large Graph Visualization.

Area of Science:

• Computer Science
• Data Visualization
• Machine Learning

Background:

• Graph layout significantly impacts information perception.
• Selecting optimal graph layouts is crucial but computationally intensive.
• Current methods rely on subjective aesthetic criteria and visual inspection.

Purpose of the Study:

• To develop a machine learning approach for large graph visualization.
• To estimate aesthetic metrics of different graph layouts efficiently.
• To introduce a novel framework for designing graph kernels.

Main Methods:

• Utilized graph kernels to compute topological similarity.
• Developed a new framework for designing graph kernels.
• Employed machine learning for layout estimation.

Main Results:

• The proposed method significantly reduces computation time compared to traditional layout calculations.
• The developed graph kernels demonstrate superior performance in both speed and accuracy.
• User studies confirmed that computed topological similarity aligns with human perceptual similarity.

Conclusions:

• Machine learning with graph kernels offers an efficient and accurate solution for large graph visualization.
• The new graph kernel framework enhances the estimation of aesthetic metrics.
• The approach effectively bridges computational analysis with human perception of graph layouts.