home
past talks
resources
conferences
|
|
|
Document Classification via Multiple Linear SVM Projections
Gary Huang
UMass
Abstract
Ensemble methods such as boosting and bagging have demonstrated effectiveness in the classification domain of machine learning, and linear support vector machines are consistently among the best-performing classifiers for textual data. In this talk, I will describe an ensemble learner that combines linear SVMs with decision trees, and present results on several text datasets. Differences and similarities to an approach that uses linear discriminant analysis (http://www.cs.ust.hk/vldb2002/VLDB2002-proceedings/papers/S19P01.pdf) will be highlighted.
Joint work with Dennis DeCoste during Summer 2004 at Yahoo! Research Labs.
Back to ML Lunch home
|