Algorithmics of Data Streams

EECS 684, Current Topics in Databases, Winter 2005

Data streams arise in many important areas in computer science including networks of small sensors, internet data monitors, and observational data about the earth and cosmos. The rate and volume of these streams pose challenges to traditional database management systems and limited-resource computing platforms generally. Fortunately, in the last 5--10 years, there have been many substantial advances in algorithms for processing data streams efficiently. Many of these algorithms, by settling for approximate answers (with provable guarantees as to quality), realize tremendous cost savings in storage space and item processing time compared with naive solutions.

In this class, I will present the necessary background in mathematics and algorithms, then students will present recent research papers in the field of data stream algorithmics. Typical results include building and maintaining standard database summaries and statistics like histograms, quantiles, and heavy hitter summaries; analysis of sensor networks. It is expected that the papers presented will cover a range of systems, experimental, and theoretical issues while fitting within the framework of data stream algorithmics.

Expected work

Each student will present a current research paper chosen in consultation with the instructor, participate in class discussion, and submit a final project.

Background and Prerequisites

The material can be approached from a variety of ways. Students should possess knowledge of database, algorithms, statistics, or mathematical modeling and approximation. Background material will be presented in class or assigned individually, as needed; students should be prepared to come up to speed quickly. Knowledge of specific database languages, like SQL, is not necessary.

Official Title EECS 684, Current Topics in Databases
Theme Algorithmics of Data Streams
Time TTh, 9-10:30, Winter, 2005
Place EECS 3437
Prerequisites EECS 484 or permission
Webpage http://www.eecs.umich.edu/~martinjs/eecs684
Instructor Assistant Professor Martin Strauss
Office, central campus 3063 East Hall
Office, north campus 2238 EECS
Email martinjs .at. umich.edu