Efficient algorithms for processing and mining large datasets

Sanjay Ranka

University of Florida, Gainesville

Thursday, May 10, 2007
Talk: 10:45 am - 11:45 am
3105 Engineering

Host: Matt Mutka

A number of scientific and medical applications require processing, transmitting and mining of large and complex data sets. Our research has been focused on developing algorithms and software for such applications.

In this talk, we will briefly describe our work on the following applications:

1. Mining metabolic networks for drug discovery: The human body at its genetic core consists of enzymes that catalyze reactions. These reactions generate compounds by consuming other compounds. Metabolic networks define the relationships between enzymes, compounds and reactions. We will describe novel computational methods for using metabolic networks in identification of target enzymes. Identification of these enzymes can accelerate the preclinical phase of the drug discovery process.
2. Grid Resource Scheduling: Many grid based E-Science applications require transfer of large files. We will describe our work on scheduling bulk file transfers, where each transfer has a start time and an end time. We have formulated the problem as a special type of multi-commodity flow problem and have developed algorithms using this formulation. Simulation results show that these algorithms improve throughput and reduce rejection rate during admission control.

If time permits, we will also present our work on algorithms for feature detection and change detection in health care data.