HOME    »    PROGRAMS/ACTIVITIES    »    Annual Thematic Program
Talk Abstract
Graph Partitioning for Dynamic, Adaptive and Multi-phase Computations

  Vipin Kumar
Department of Computer Science and Engineering
University of Minnesota
4-192 EE/CS Building
Minneapolis, MN 55455
kumar@cs.umn.edu


Algorithms that find good partitionings of highly unstructured graphs are critical in developing efficient algorithms for problems in a variety of domains such as scientific simulations that require solution to large sparse linear systems, VLSI design, and data mining. Even though this problem is NP-hard, efficient multi-level algorithms have been developed that can find good partitioning of static irregular meshes. The problem of graph partitioning becomes a lot more challenging when the graph is dynamically evolving (e.g., in adaptive computations), or if computation in multiple phases needs to be balanced simultaneously. This talk will discuss these challenges, and then describe some of our recent research in addressing them.

Joint work with Kirk Schloegel and George Karypis.


Material from IMA Talk

Back to Workshop Schedule

Back to Low-speed Combustion

1999-2000 Reactive Flow and Transport Phenomena

Go