Date: Wednesday, 9 April 2008, 6:30 PM
Location: SAP LABS, Building D, 3410 Hillview Avenue, Palo Alto, CA (Google Maps | Yahoo! Maps | Mapquest)
Cost: Free and open to all who wish to attend, but membership is only $10/year.

Topic

Many problems in machine learning are cast as constrained optimization problems. The talk focuses on efficient algorithms for learning tasks which are cast as optimization problems subject to L1 and box constraints. The end result are typically sparse and accurate models. We start with an overview of existing projection algorithms onto the simplex. We then describe a linear time projection for dense input spaces. Finally, we describe a new efficient projection algorithm for very high dimensional spaces. We demonstrate the merits of the algorithm in experiments with image and large scale text classification.

About the Speaker

Yoram Singer is a senior research scientist at Google Inc. From 1999 through 2007 he was an associate professor of computer science at the Hebrew University, Jerusalem, Israel. From 1995 through 1999 he was a member of the technical staff at AT&T Research. His work focuses on the design, analysis, and implementation of machine learning algorithms.

Slides from the Presentation [PDF]

Back to the DM SIG page