Accurate Visual Display of Quantitative Quantum Physics Data
Henan Zhao Ph.D. Defense
Meeting number (access code): 2626 617 2999
Meeting password: x2EJvriJX82
Committee: Drs. Marc Olano (Chair), Jian Chen (Co-Chair), Garnett Bryant, Victoria Interrante, Matthew Pelton, Judith Terrill
The rise of data-driven quantum physics has motivated the use of data visualization to communicate data and explore insights to users. In contrast to raw data, visual representations are more expressive and can communicate more information. The pervasive and effective uses of visualization are supposed to improve the scientific process and let users easily inspect data from experiments and simulations. A critical feature is perceptually accurate exploration that provides accurate interpretations of patterns in response to users' input. This form of representation is crucial for viewers searching for patterns or asking "what-if" questions. Due to the increasing complexity (size and density) of modern quantum physics simulation, there is a lack of design guidelines for quantum physicists to explore complex simulations.
The first contribution in this dissertation work is a novel visual representation called SplitVectors that supports conveying large-range magnitude in quantum physics simulations accurately. This is an important problem as many simulation results have large-range data. We hypothesized that in many cases such data could be more accurately analyzed when projected into bivariate glyphs with scientific notations rather than direct depictions. Our design dramatically improved the visual discrimination of vectors in space and achieved up-to 10 times better accuracy.
An often overlooked element of interactive data exploration is the benefits of global gist of the scene which is also pre-attentive based on most recent vision science theory. While computational and data processing capabilities have increased over time, humans have cognitive limits. Yet, it is a debate how we can still perceive the rich physical world and selectively attend to a small set of important items. We recognize these new advances and understand the use of global structure as a way to design visualizations to minimize human computing cost. The second contribution of this dissertation work is to reveal that visual marks (color, texture, and length) vary its power and coloring and texture are most useful to guide quantum physicists’ tasks of seeing from a large chuck of items. Our eye-tracking experiment revealed two intriguing strategies that viewers used when searching for items from hundreds of similar items. Viewers used global distribution (gist of the scene) even without knowing the individual items and were able to restrict eye movement to the task-relevant regions of interest - we called them drillers. Drillers had the least trial errors, and categorical coloring allowed more drillers thus made a more effective display.
Equally challenging is that the resulting quantum physics datasets are rich and provide many opportunities for discovery, but supporting feature explorations across highly dense three-dimensional (3D) simulations results is extremely challenging. The third contribution of this work is that we apply searching and comparison-oriented strategies and couple two- dimensional (2D) displays and 3D spatial data in a “world-in-world” environment to optimize interaction across the entire set of interactive queries of many tasks. Our case studies yield insights that could lead to improved design for integrating 2D and 3D solutions for complex multi-scale data analyses. We demonstrate the use of our concepts in both desktop and immersive virtual environment settings.