|
|
|
sTrustr
Solves convex nonlinear programming problems.
Special handling of separable problems.
Main features
-
sTrustr implements a
structured trust region algorithm by
A. R. Conn, Nick Gould, A. Sartenaer, Ph. L. Toint;
Convergence Properties of Minimization Algorithms For Convex
Constraints Using A Structured Trust Region, July 4, 1995.
The algorithm is
combined with an initial
trust region radius algorithm, ITRR, by Sartenaer, 1995.
-
The algorithm treats nonlinear constraints, but will fail for non-convex
problems.
Use the solvers conSolve and nlpSolve in the Tomlab Base Module,
MINOS in Tomlab /MINOS,
or NPSOL in Tomlab /SOL
for such problems.
-
Special treatment of partially separable functions, where
f(x) = sum (i=1:M) f_i (x).
-
Quasi-Newton BFGS or DFP is used for the Quasi-Newton update,
if second order information is not present
or the analytical Hessian is not used.
|
|