1 of 1

A Comparative Study of the Perceptual Sensitivity of Topological Visualizations to Feature Variations

Scientific Achievement

ORNL researchers detailed their study observing the sensitivity of topological data analysis-based visualization methods, which measure how well changes in data structure can be perceived. Their work provided a first-of-its-kind empirical evaluation of the efficacy of a range of feature representations against a reference of color mapping in performing a visual comparison task for a scalar field, or a field where there is a single number associated with every point in space.

Significance and Impact

These results display which topology-based tools are most effective for different data and task scenarios in conveying topological variations, an example of which would be tracking the eye of a hurricane. The study showed color maps performed overall well in accuracy and sensitivity but isocontours did not. While Reeb graphs and persistence diagrams demonstrated precise utility, no one visualization outperformed the rest.

Sensitivity analysis results for a color map [column (a)] and topological visualizations [columns (b)-(d)]. Bar charts depict the percentage of correctly answered trials for the positional data variation (center row) and amplitude data variation (bottom row). Steeper slope of a red line in bar charts indicates the higher sensitivity to data variations. Based on our study, no single visualization type effectively conveyed both position and amplitude variation in the data, which necessitates investigation of new visualization types.

Technical Approach

Researchers ran the study as a human-subject study that evaluated the perception of data features characterized by Gaussian signals, which are signals with a probability density function equal to the normal distribution.
The study measured how effectively each visualization technique portrays variations of data features arising from the position and amplitude variation of a mixture of Gaussians.

PI(s)/Facility Lead(s): Paul Rosen, University of Utah; Tushar Athawale, ORNL

Collaborating Institutions: University of South Florida, University of Utah, ORNL

ASCR Program: RAPIDS-2 SciDAC

ASCR PM: Kalyan Perumalla, Hal Finkel

Publication(s) for this work: Athawale, Tushar M, et al., “A Comparative Study of the Perceptual Sensitivity of Topological Visualizations to Feature Variations,” IEEE Transactions on Visualization and Computer Graphics, Vol. 30, No. 1 (2024): Jan.

DOI: 10.1109/TVCG.2023.3326592.

LOCAL LAB POC: Ken Moreland and Scott Klasky

TALKING POINTS:

ORNL researchers detailed their study on the sensitivity of topological data analysis-based visualization methods. This measures how well changes in data structure can be perceived.
Their work provided a first-of-its-kind empirical evaluation of the efficacy of isocontours, Reeb graphs, and persistence diagrams against a reference of color mapping in performing a visual comparison task for scalar fields.
These results take an important step toward understanding which topology-based tools are best for different data and task scenarios in conveying topological variations as compared to conventional color mapping.

METADATA:

Name of the associated awarded project:

PI name(s): Paul Rosen, University of Utah

Local lab POC: Tushar Athawale, ORNL

Name of the program manager: Hal Finkel

CITATIONS:

The notes section should also include full citations (including the DOI) to any important and associated publications, datasets or code developed as part this work.

Publication(s) for this work: Athawale, Tushar M, et al., “A Comparative Study of the Perceptual Sensitivity of Topological Visualizations to Feature Variations,” IEEE Transactions on Visualization and Computer Graphics, Vol. 30, No. 1 (2024): Jan.
DOI:10.1109/TVCG.2023.3326592

AWARDS:

REPRODUCIBILITY:

BACKGROUND AND CONTEXT INFORMATION: