Campuses:

Robust Multi-Scale Morse Analysis for Scientific Data Analysis and Exploration

Friday, October 28, 2005 - 1:25pm - 2:25pm
Valerio Pascucci (Lawrence Livermore National Laboratory)
In recent years topological analysis has emerged as a fundamental tool for extracting intrinsic geometric features present in scientific data. In this talk I will discuss the challenge of developing algorithms that enable the practical use Morse theory in the analysis in real scientific data, in other words algorithms that are robust, comprehensive, and multi-scale. Robust algorithms avoid the numerical instabilities that are particularly problematic when dealing with noisy data. A comprehensive analysis guarantees reliable data understanding and exploration. Multi-scale representations allow isolating features of interest at different resolutions.

I will illustrate in detail how these concepts can be translated systematically into efficient algorithms and data structures. The resulting approach is used for constructing Morse-Smale complexes, Contour trees, and Jacobi sets that are used in user interfaces components for general purpose data exploration. The same techniques are also used for specialized data analysis tools for time tracking of particles in combustion simulations, for the characterization of bubbles and spikes in Rayleigh-Taylor instabilities, and for the reconstruction of complex channel structures in porous media.

I will provide a live demonstration of the approach with a simple visualization tool with linked views coordinating the presentation of topology with traditional visualization techniques. More details regarding this work can be found at http://www.pascucci.org.