Wednesday, October 17, 2012
CSCAMM Seminar - Prof. Andre Tits - Constraint Reduction in Interior-Point Methods for Linear and Convex Optimization
Description:
Constraint reduction is a technique by which each search direction is computed based only on a small subset of the inequality constraints (when the problem is expressed in standard dual form), containing those deemed most likely to be active at the solution. A dramatic reduction in computing time may result for severely imbalanced problems.
In this talk, we survey developments made at UMPC over the past few years, in a group led by Dianne O'Leary and the presenter. The power of constraint reduction is demonstrated on classes of randomly generated problems and on real-world applications. Numerical comparison with both simplex and "unreduced" interior point is reported.
Note: This talk has a significant overlap with a talk given by the author in the UMCP NA Seminar Series in April 2012.
{Will appear in FYI on Oct 17, 2012
Start Time:
2:00 PM
End Time:
3:00 PM
Room:
4122
Common Location Name:
Computer Science Instructional Center
Web Address:
www.cscamm.umd.edu/seminars/fall12/
Other Contact Information:
Ann Ekechukwu
CSCAMM
+1 301 405 0652
ekechann@umd.edu
www.cscamm.umd.edu/index.aspx
