Colloquium
Hal Daume
University of Maryland
Wednesday, March 21, 2012
1230 WEB
Refreshments 3:20 p.m.
Lecture 3:40 p.m.
Host: Ellen Riloff
Title: Complex Predictions Need Not Be Slow
Hal Daume
University of Maryland
Wednesday, March 21, 2012
1230 WEB
Refreshments 3:20 p.m.
Lecture 3:40 p.m.
Host: Ellen Riloff
Title: Complex Predictions Need Not Be Slow
Abstract
Classic algorithms for predicting structured data (eg., graphs, trees, etc.) or rankings rely on expensive (sometimes intractable) inference at test time. In this talk, I'll discuss several recent approaches that enable computationally efficient (eg., linear-time) prediction at test time. These approaches fall in the category of learning algorithms that explicitly optimize a speed/accuracy trade-off, or optimize accuracy under a fixed computational budget. This is joint work with: Jason Eisner, Lise Getoor, Jiarong Jiang, He He, Adam Teichert, Tim Vieira, Jay Pujara and Lidan Wang.
Classic algorithms for predicting structured data (eg., graphs, trees, etc.) or rankings rely on expensive (sometimes intractable) inference at test time. In this talk, I'll discuss several recent approaches that enable computationally efficient (eg., linear-time) prediction at test time. These approaches fall in the category of learning algorithms that explicitly optimize a speed/accuracy trade-off, or optimize accuracy under a fixed computational budget. This is joint work with: Jason Eisner, Lise Getoor, Jiarong Jiang, He He, Adam Teichert, Tim Vieira, Jay Pujara and Lidan Wang.