University of Pennsylvania
Wu and Chen Auditorium
101 Levine Building
3330 Walnut Street
Philadelphia, PA

Tuesday, October 14, 2008

 

The Northeast DB/IR Day is a forum for bringing database, machine learning, and information retrieval researchers and students from the Northeast region of the US – including both academic and research institutions.  DB/IR Day aims to promote discussion across the database, learning, and information retrieval communities, and also to promote interactions across organizations and universities.  It is held semi-annually, rotating among the major research institutions in the general Northeast US region.  This Fall it is hosted by the University of Pennsylvania's database group.

Attendance is free, but we request you register ahead of time so we can ensure sufficient boxed lunches and refreshments.  The program (shown below) consists of three keynote speakers, as well as student posters.  We have a limited number of poster boards and easels, and request that you register these in advance:  they will be given out on a first-registered, first-served basis.

Preliminary Program

9:00 - 9:50	Registration, coffee, informal discussions (Levine lobby)
9:50 - 10:00	Opening remarks
10:00 - 11:00	Keynote speaker:  Ken Ross, Columbia University (Levine 101) How Architectural Knowledge Influences Algorithm Design for Databases Modern computer architectures contain a variety of low-level features designed to improve performance. Examples include SIMD units, out-of-order execution, overlapping of memory requests, chips with multiple processing cores, etc. In this talk I will argue that knowledge of the underlying architecture is necessary if one wants to design efficient algorithms for modern machines. I will illustrate this contention with two examples. First, I will show how a variant of cuckoo hashing can outperform traditional hashing in terms of both space and probe time on common commodity architectures. Second, I will show how one can implement aggregation on a highly parallel architecture. Real-time sampling is used to estimate performance characteristics and adaptively select an appropriate algorithm within an aggregation operator.
11:10 - 12:10	Database & Information Retrieval research overviews: a description of the work going on in some of the major DB/IR centers in the region. Presenters: Sihem Amer-Yahia, Yahoo! Research; Dennis Shasha, NYU; Divesh Srivastava, AT&T Labs-Research; Hila Becker, Columbia; Boon Thau Loo, Penn.
12:10 - 1:30	Lunch (obtain boxed lunches in Levine 307; eat in lobby or outdoors)
1:30 - 2:30	Poster presentations (Levine lobby)
2:30 - 3:00	Coffee and discussion break (Levine 101)
3:00 - 4:00	Keynote speaker:  HV Jagadish, University of Michigan (Levine 101) Making Database Systems Usable In this talk, I will examine why database systems today are so difficult to use and identify a set of five pain points. I will describe the work we have been doing in my research group on database usability, and propose a research agenda to address these pain points. In particular, we propose a presentation data model as a formal layer above the logical and physical data model layers in a database system.
4:00 - 4:15	Student poster awards and closing remarks (Levine 101)

Registration

Registration is free. Please sign up if you are to attend, and also if you want to present a poster.

Location

For the Fall, DB/IR Day is being held at the home of the University of Pennsylvania's Computer and Information Science Department, Levine Hall.  Please see below for a detailed map:

View Larger Map

 

Transportation: Amtrak stops at 30th Street Station in Philadelphia, which is only a few blocks from the Levine Building (30th and Market, vs. 33rd and Walnut). The SEPTA R1 train from the Airport stops right outside the Penn campus (University City stop). The Market-Frankfort El (subway) stops at 34th and Market, approximately 3 blocks from our building.