SQP

Nonlinear programming

SQP uses an implementation of Powell's successive quadratic programming algorithm and is aimed specifically at large, sparse nonlinear programs. It solves the quadratic programming subproblems by using a sparsity-exploiting reduced gradient method. Sparse data structures are used for the constraint Jacobian, and there is an option to represent the approximate Hessian as a small set of vectors using a limited memory updating scheme.

SQP requires the same user-supplied subroutines as GRG2 and has similar subroutine and data file interfaces. The entry describing GRG2 contains more details.

SQP is written in ANSI Fortran. Machine dependencies are relegated to the subroutine INITLZ, which defines three machine-dependent constants.

Need more info?

Contact:

Prof. Leon Lasdon 
MSIS Department 
College of Business Administration 
The University of Texas at Austin 
Austin, TX 78712-1175 
Phone: (512) 471-9433
http://www.optimalmethods.com

References:

Y. Fan, S. Sarkar, and L. Lasdon, Experiments with successive quadratic programming algorithms, J. Optim. Theory Appl. 56 (1988), pp. 359--383.

D. Mahidhara and L. Lasdon, An SQP algorithm for large sparse nonlinear programs, Working Paper, MSIS Department, College of Business Administration, The University of Texas at Austin, 1991.

[ Optimization Software Guide | OTC Home Page | NEOS Server | NEOS Guide ]