The rapid growth of data size and computing power is a constant motivation for the visualization algorithms to adapt and advance. In the presence of large-scale data, it remains a challenge to generate visualizations which can guide the user to explore the data interactively and through meaningful queries.

My talk will focus on a visual analytic framework designed to address this problem. The framework uses suitable statistical or geometric measures to quantify importance with respect to the goal of current analysis, performs importance-based data reduction, and retains data summaries which can be repeatedly used to answer user queries and generate meaningful visualizations. I will present two instances of this framework with case studies involving simulation data, either volumetric or geometric in nature, widely used in various scientific disciplines.