This algorithm is based on the paper
Ferreira, P. S., Karas, E. W., Sachine, M., & Sobral, F. N. C. (2017). Global convergence of a derivative-free inexact restoration filter algorithm for nonlinear programming. Optimization, 66(2), 271–292.
FIRD
is able to solve general derivative-free constrained
optimization problems in which the derivatives of the objective
function are not available.
In the provided version, FIRD
uses a modified version of the
algorithm TRDF
for building and
updating the trust region models. To solve the nonlinear trust-region
problems, it uses ALGENCAN
.
To install and build ALGENCAN
library
-
Unpack it and, in the main folder, e.g
$ALGSRC
, simply typemake
in order to generate$ALGSRC/lib/libalgencan.a
file;
-
Download and unpack
FIRD
; -
In the main subdirectory, edit file
Makefile
with the path toALGENCAN
librarySOLVERLIB = $ALGSRC/lib
where
$ALGSRC
is the name of the full path whereALGENCAN
has been installed; -
Type
make fird
to generate test problem HS14 from the Hock-Schittkowski collection. The executable file will be generated in
/bin
subdirectory. Or, typemake fird PROBLEM=tests/examples/toyprob.f90
for a small example.
-
Run it
./bin/fird
FIRD
can be modified in order to use different solvers for
- Feasibility phase
- Construction and updating models (optimality phase)
- Solving nonlinear programming trust-region subproblems (optimality phase)
The manual for advanced information can be generated with the command
make doc