BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20111117T213000Z
DTEND:20111117T220000Z
LOCATION:TCC 305
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Eigensolvers are important tools for analyzing and mining useful=0A  information from scale-free graphs.  Such graphs are used in many=0A  applications and can be extremely large.  Unfortunately, existing=0A  parallel eigensolvers do not scale well for these graphs due to the=0A  high communication overhead in the parallel matrix-vector=0A  multiplication (MatVec).  We develop a MatVec algorithm based on 2D=0A  edge partitioning that significantly reduces the communication costs=0A  and embed it into a popular eigensolver library.  We demonstrate=0A  that the enhanced eigensolver can attain two orders of magnitude=0A  performance improvement compared to the original on a state-of-art=0A  massively parallel machine.  We illustrate the performance of the=0A  embedded MatVec by computing eigenvalues of a scale-free graph with=0A  300 million vertices and 5 billion edges, the largest scale-free=0A  graph analyzed by any in-memory parallel eigensolver, to the best of=0A  our knowledge.
SUMMARY:A Scalable Eigensolver for Large Scale-Free Graphs Using 2D Partitioning
PRIORITY:3
END:VEVENT
END:VCALENDAR