This project addresses fundamental challenges in incorporating and conveying uncertainty in the process of data analysis and visualization. In order to compute uncertainty, the project aims to develop a general model for uncertainty analysis that is independent of the visualization method and application domain. An important and novel aspect of the technical approach is the division of uncertainty into high and low conceptual levels, with separate but complementary methods of analysis for each level. The ultimate goal is to allow assessment of uncertainty at the level of tasks and queries, aided by effective visualization. The project focuses on three principal research problems: (1) development of a hybrid probability-possibility uncertainty analysis framework for representing low-level computational uncertainty, along with methods for displaying this uncertainty in various visual modalities; (2) formulation of a fuzzy analysis for representing high-level, task-related uncertainty that handles human input and visual/perceptual uncertainty, while bridging the gap between low-level uncertainty and high-level uncertainty; and (3) investigation into ways to visually display the evolution of uncertainty in computation, enabling uncertainty navigation, in which exploration and modification of uncertainty can occur in the same context. The resulting framework is expected to effectively enable verifiable visualization of uncertainty in data analysis.
This project draws from many fields of research outside of visualization, including management of uncertainty, fuzzy logic, information theory, data analysis, simulation, computer vision, human-computer interaction, computer graphics and high performance computing and its potential impact extends to these areas and beyond. The ultimate goal of verifiable visualization is beneficial to visualization and visual analytics, but also facilitates the adoption of visualization in other fields, such as medical imaging, computational biology, and visual analytics to name a few. The project results will be disseminated to the visualization community and beyond through annual conferences, workshops, and tutorials, and also through the project website (http://vis.cs.ucdavis.edu/NSF/IIS1320229), which will include project status updates and deliverables such as images, videos, and prototype software. Complementing the proposed research is an educational agenda, consisting of integration of research results into teaching, arrangement of summer internships for participating students at the collaborating scientists' laboratories, and involvement of graduate and undergraduate students in research.
