Transformed Hessian methods for large-scale constrained optimization

Conference · Fri Dec 30 23:00:00 EST 1994

OSTI ID:36060

Gill, P; Murray, W; Saunders, M

We define the transformed Hessian and discuss its role in sequential quadratic programming (SQP) methods for constrained optimization. In large-scale optimization, nonlinear inequality constraints are usually converted to equalities by adding bounded slack variables. Provided the proper sparse-matrix techniques are used, there is no significant penalty with the increased problem size. We identify structure in the transformed Hessian that is present when there are slack variables and describe how this structure may be exploited in a quasi-Newton method. Numerical results from the large-scale SQP code SNOPT will be presented.

• OSTI ID: 36060
• Report Number(s): CONF-9408161--
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
ACCURACY
ALGORITHMS
CONVERGENCE
MATRICES
NEWTON METHOD
NUMERICAL SOLUTION
OPTIMIZATION
VARIATIONAL METHODS