BEGIN:VCALENDAR
PRODID:-//University of Illinois//Web Services Calendar//EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
DTSTAMP:20120214T133803Z
DTSTART;TZID=America/Chicago:20091105T160000
DTEND;TZID=America/Chicago:20091105T160000
SUMMARY:Statistics Seminar: ``Finding Significant Large-Average Submatric
 es in High Dimensional Data" by Andrew Nobel\, UNC
CREATED:20090811T120000Z
DESCRIPTION:Abstract: Exploratory analysis of high dimensional data often
  begins withindependent clusteringof samples and variables\, yielding a 
 partition of the available datamatrix into disjointrow-column blocks (su
 bmatrices). Of particular interest in practiceare submatriceswhose entri
 es are large on average.  In conjunction with clinical andfunctional ann
 otation\,large average submatrices are frequently the starting point for
 subsequent analyses\,such as the identification of genetic pathways and 
 new diseasesubtypes in the study ofgene expression data.This talk descri
 bes a simple algorithm\, belonging to the generalcategory of biclusterin
 gmethods\, for identifying large average submatrices in high dimensional
 data. Like otherbiclustering methods\, the algorithm improves on indepen
 dent samplevariable clusteringin several respects: the submatrices it id
 entifies can overlap andthey need not cover theentire data matrix (featu
 res that better reflect the underlyingstructure of many problems)\,and t
 he inclusion ofsamples and variables in a submatrix does not depend on t
 heirexpression values outside thesubmatrix. The algorithm seeks to maxim
 ize a simple measure ofstatistical significance\, whichalso provides an 
 objective basis for comparing and selecting amongsubmatrices of differen
 t sizesand average intensities.  I will discuss the applications of the 
 algorithm to arecent gene-expression based cancer study\, and will provi
 de a detailedcomparisonof its performance with several other biclusterin
 g method\, includingits applicationto semi-supervised classification.
LAST-MODIFIED:20091022T120000Z
LOCATION:(ROOM CHANGE) 122 Illini Hall
CATEGORIES:Seminar
CONTACT:3-2167
ORGANIZER:office@illinois.edu
URL:http://illinois.edu/calendar/detail/1439?key=2000010120000101126566
UID:126566@illinois.edu
END:VEVENT
END:VCALENDAR