The Solver for Local Nonlinear Optimization Problems version 1.1, June, 1997 MSFWin32 FORTRAN source library for use with MS PowerStation FORTRAN Optimizing Compiler 1.0 by Alexei V. Kuntsevich & Franz Kappel Karl-Franzens University of Graz Institute for Mathematics Heinrichstr. 36 A-8010 Graz (Austria) E-mail: franz.kappel@kfunigraz.ac.at The program SolvOpt (Solver for local optimization problems) is concerned with minimization resp. maximization of nonlinear, possibly non-smooth objective functions and with the solution of nonlinear programming problems taking into account constraints by the so-called method of exact penalization. SolvOpt can solve the following nonlinear optimization problems: Type Problem Unconstrained min{f(x): x in R(n)} Optimization max{f(x): x in R(n)} Constrained min{f(x): g(x)<=0, h(x)=0, x in R(n)} Optimization The routine SOLVOPT requires that the user supplies at least a subroutine containing the code which computes the values of the objective function at a given point. Unless the user also provides a subroutine containing the code for computing the (sub)gradient of the objective function, these (sub)gradients are calculated by the program itself using finite differences. For the constrained problem, the user is required to supply additionally at least a routine containing the code which computes the maximal residual for a set of constraints at a given point. For more information see the Tutorial part of the manual on SolvOpt. ------------ * ------------ C o n t e n t s solvopt.for is the optimization routine SolvOpt (SOLver for local nonlinear OPTimization problems), dualshl0.for is a sample problem file illustrating the case of a priori chosen penalty coefficient. It also provides an example of passing the user's parameters by use of entry points to a subroutine. dualshl1.f(or) is a sample problem file illustrating the case of a priori chosen penalty coefficient. It also provides an example of passing the user's parameters by use of FORTRAN common blocks. dualshl2.f(or) is a sample problem file illustrating the case of a priori chosen penalty coefficient. It also provides an example of the use of the finite difference approximation of the gradients. dualcnst.f(or) is a sample problem file illustrating the case of solving an NLP problem by use of penalization. The penalty coefficient is chosen automatically by the solver. dualcnt2.for is a sample problem file illustrating the case of solving an NLP problem by use of penalization. The penalty coefficient is chosen automatically by the solver. In the case, the gradients are approximated by the finite differences. rosenbr.for is a sample unconstrained problem file containing the routines required for minimization of Rosenbrock function. readme is this file. shor.for is a sample nonsmooth unconstrained problem file containing the routines required for minimization of Shor piece-wise quadratic function. messages.inc is an include file to compile with solvopt.f(or) which contains definitions of warning and error messages. soptions.for returns the default values for the optional parameters. ------------ * ------------ The libraries (for use with MS FORTRAN) and (for use with HP FORTRAN-90) contain different source codes. !!! PLEASE, DOWNLOAD THE APPROPRIATE ARCHIVE FILE !!!