Saturday, May 10

MS14
Graph Coloring for the Computation of Sparse Jacobians and Hessians

3:00 PM - 5:00 PM
Room: Franklin - CC

Sparse Jacobians and Hessians can be computed more efficiently using automatic differentiation (and, sometimes, finite differences) through compression techniques based on the sparsity pattern. This session covers techniques for detecting spasity patterns, models for graph coloring, and serial and parallel methods for

Organizer: Paul D. Hovland
Argonne National Laboratory

3:00-3:25 Pattern Graphs for Sparse Matrices

Shahadat Hossain, University of Lethbridge, Canada

3:30-3:55 Sparsity Detection for Jacobians and Hessians in ADOL-C

Andrea Walther, Technische Universitaet Dresden, Germany

4:00-4:25 How to Compute a Jacobian More Efficiently or Coloring Highly Structured Graphs

Dan Cranston, DIMACS

4:30-4:55 Acyclic and Star Coloiring Algorithms for Hessian Computation

Assefaw H. Gebremedhin, Purdue University

Cancelled 4:00-4:25 Coloring Hierachical Derivative Matrices

Paul D. Hovland, Argonne National Laboratory