Tuesday, May 11
MS17
Optimization Challenges in Data Mining
10:45 AM-12:45 PM
Room: Georgia 8
This minisymposium examines optimization problems found in data
mining. Large quantities of data have been captured by computerized
applications. Robust methods for prediction, clustering and
visualization/browsing are needed to allow users to extract useful
information from this data. Many of these problems can be framed as
optimization problems whose difficulty is compounded when applied to
massive high-dimensional databases. The speakers will present recent
research on optimization based approaches to the modeling and
solution of data mining problems.
Organizer: Kristin P. Bennett
Rensselaer Polytechnic Institute
- 10:45-11:10 Scaling Clustering Algorithms to Large Databases
- Paul Bradley, Usama Fayyad, and Cory Reina, Microsoft Research
- 11:15-11:40 Divide
and Conquer: An Algorithm for Minimizing the Training Sample
Misclassification Cost in Two-Group Classification
- Antonie Stam, University of Georgia; and A. Pedro Duarte
Silva, Universidade Catolica Portuguesa, Porto, Portugal
- 11:45-12:10 Linear
Programs for Accuracy Control in Classification and Regression
- Bernhard Schölkopf and Alex
Smola, Australian National University, Canberra and GMD First,
Berlin, Germany
- 12:15-12:40 Combining Supervised and Unsupervised Learning
Using Transduction
- Kristin Bennett, Organizer
MMD, 5/3/99