Scientific simulations generate a massive amount of data, from initial conditions or parameter settings at one end to multi-dimensional probability distributions, to random fields or ensemble collections at the other end. These datasets are often high-dimensional, time-varying, and multifaceted. Visualization plays an indispensable role in the chain of knowledge discovery from big data. A significant problem faced by domain scientists is the robust treatment of uncertainty. The uncertainty stems from computational models, training data decimation or abstraction, and even the process of visualization. Visualizing uncertainty is important as it presents more comprehensive and accurate information for human understanding, model comparison, and simulation steering. However, it is difficult to quantify the source of uncertainty, investigate its possible propagation, and represent it appropriately for visual comprehension.
CICS research topics in uncertainty visualization include statistical and probabilistic approaches for uncertainty identification and quantification, visual representations for encoding and summarization of uncertainty information, analytical interfaces for visual interaction, verification, and validation, and deep learning solutions for sensitivity analysis, uncertainty mitigation, and predictive analytics.
Faculty involved in this research area include Chaoli Wang.