Search Results

Network data is widely used to represent complex systems across domains, enabling the discovery of structural patterns and relationships. However, as graphs grow in size, visualizing them effectively becomes increasingly challenging. BioFabric offers a compact representation in which nodes are drawn as horizontal lines and edges as vertical lines, eliminating edge crossings by design and supporting the identification of meaningful motifs such as staircases. Despite these advantages, the readability of BioFabric layouts depends critically on the ordering of nodes and edges, making both computation and visualization difficult to scale to high cardinality data. To address these challenges, we explore progressive visualization, a paradigm that allows users to inspect increasingly accurate partial results as data processing continues. Applying progressive visualization to BioFabric introduces a key difficulty: intermediate layouts may contain correct subsets of the data yet remain visually misleading if their ordering diverges significantly from the final layout. We frame this problem as one of ordering convergence to investigate how different sampling strategies affect the evolution of the layout. Specifically, we design a benchmark to evaluate how quickly intermediate visualizations approximate the final BioFabric ordering using measures such as Kendall-τ, and analyze the impact of progressive computation on the emergence of recognizable patterns, such as staircases. Our results provide a first characterization of how graph topology and progressive data chunking strategies influence convergence and layout accuracy. We derive best practices for designing progressive BioFabric systems that become informative before the full computation is complete. Two case studies show the benefits of progressively sampling BioFabric.