Using Indicators in Finite Termination Procedures
Pamela Williams
Sandia National Labs, Livermore, California
Thursday, August 1, 2002 15:00-16:00,
Room 145, NIST North (820)
Gaithersburg
Thursday, August 1, 2002 13:00-14:00,
Room 4511
Boulder
Abstract:
Indicator functions identify the active set at the solution of a
constrained optimization problem. Early identification of the active set
may result in computational savings for interior-point algorithms for linear
programming problems. First, we discuss the merits and convergence
properties of three commonly used indicators in practice. Next, we
investigate the role of indicators in finite termination procedures. In a
finite number of steps, finite termination procedures advance from an
interior-point iterate which is sufficiently close to the solution set to a
point on the optimal face. Empirical evidence supports the use of finite
termination procedures to reduce computational work and to generate highly
accurate solutions.
Contact: A. J. KearsleyNote: Visitors from outside NIST must contact
Robin Bickel; (301) 975-3668;
at least 24 hours in advance.
|
