Seminars & Colloquia
Oak Ridge National Laboratory
"Towards Data Analytics at Petascale: Finding the Dots, Connecting the Dots, Understanding the Dots"
Thursday March 22, 2007 10:00 AM
Location: 3211, EB II NCSU Centennial Campus
(Visitor parking instructions)
In this talk, I will highlight some of the needs for and our progress towards petascale data analytics that would enable application scientists to find informative features (identifying the dots) and to link them (connecting the dots) to formulate fundamental principles governing complex natural phenomena (understanding the dots). The difficulty lies in dealing with the millions of components and their billions of interactions on different spatial and temporal scales. Existing approaches are challenged by the curses of dimensionality, computational intractability, and noise. This talk will focus on how we could break these curses by exploiting their blessings through advanced theory, novel algorithms and their scalable implementations. Specifically, I will present our pioneering research in (a) fixed parameter tractability theory, (b) distributed and streaming data mining algorithms, and (c) an infrastructure for interactive and transparent parallel statistical computing with R. The application of these technologies to large-scale problems in astrophysics, climate, and biology will be presented.
Host: Xiaosong Ma, Computer Science