Beginning in 1983, he developed sparse matrix algorithms for massive analysis of variance and regression problems (an area at the intersection of graph theory, numerical mathematics, and statistics). With the advent of parallel computers in the mid 1980's, he pioneered some of the first parallel algorithms for data analysis. His later contributions include search methods for massive classes of hierarchical models (related to Bayes Nets), feature extraction for large data series, and more recently fast clustering and dimension reduction methods for distributed data. In addition to developing new analysis and visualization algorithms, his work also includes the application of a variety of mathematical and statistical methods in a number of areas including biology, chemical kinetics, computer networks, neutron scattering, combustion, dose measurement, climate, astrophysics, finance, air traffic, and various contamination and remediation settings.
