Visualizing the uncertainty induced by graph layout algorithms

Given a graph structure, different layout algorithms (even different settings of the same algorithm) usually result in different arrangements of vertices, and each layout may reflect certain aspects/parts of the graph more accurately than others. Thus, for high-level graph analysis tasks that rely on the overall arrangement of vertices, drawing conclusions only from one layout is risky. To alleviate the risk, we propose an ensemble framework to capture the commonalities and differences among possible layouts, and help users obtain a comprehensive view of the structure patterns. We leverage a set of layouts that represents the distribution of algorithm outputs. Then, visual features are extracted and analyzed based on various measures of visual similarity. Our framework supports users to analyze individual layouts in the context of the distribution, so that users can quickly identify structures of interest and discover patterns more accurately and comprehensively. We demonstrate the effectiveness of our framework by applying it to three datasets.

2017 IEEE Pacific Visualization Symposium (PacificVis)