Wednesday, July 13

MS38
Data Mining using Matrices and Graphs - Part I of II

10:30 AM - 12:30 PM
Room: Eglinton and Winton - 2nd Floor

For Part II, see MS49

The fields of data mining and machine learning are increasingly adapting methodologies and algorithms from advanced matrix computations, graph algorithms, and mathematical programming. These sessions present some recent work.

Session 1 focus on matrix/mathematical approaches using matrix decompositions, discrimant analysis, spectral embeddings, FFT and level sets method, with applications in image processing, fingerprint classification, information retrievl and DNA microarray data clustering.

Session 2 focus on graph algorithms, searching, ordering, ranking, and contraction, with applications on web link analysis, information discovery on complex networks, sparse matrix envelope reduction, and data clustering.

Organizer: Chris Ding
Lawrence Berkeley National Laboratory
Haesun Park
Georgia Institute of Technology
Hongyuan Zha
Pennsylvania State University

10:30-10:55 2-dimensional Singular Value Decomposition for 2D maps and Images

Chris Ding, Lawrence Berkeley National Laboratory; Jieping Ye, University of Minnesota

11:00-11:25 Equivalence of Several Two-stage Methods for Linear Discriminant Analysis

Peg Howland, Utah State University

11:30-11:55 Latent Semantic Analysis, Bipartite Graphs and Laplacian Embeddings

Bruce Hendrickson, Sandia National Laboratories

12:00-12:25 Dynamic Cluster Formation Using Level Sets

Andy M. Yip, University of California, Los Angeles; Tony F. Chan, National Science Foundation; Chris Ding, Lawrence Berkeley National Laboratory

Cancelled 11:00-11:25 Fingerprint Classification Using Fast Fourier Transform and Nonlinear Discriminant Analysis

Haesun Park, Georgia Institute of Technology