ACRS - A Controlled Random Search algorithm

How to use the derivative-free optimizer ACRS for bound constrained global optimization

The package provides a FORTRAN90, C, PYTHON and MATLAB version of the code.

1. Gunzip and untar the archive in a folder on your computer by issuing in a directory of your choice (ex. curdir) the command

   $> tar -xvf ACRS.tar.gz

• FORTRAN VERSION OF THE CODE:

  Edit file curdir/F90_src/problem.f90 to define your own objective function.

  In particular, modify the subroutines

  setdim : which sets problem dimension

  setbounds : which sets upper and lower bounds on the variables

  funct : which defines the objective function

• C VERSION OF THE CODE:

  Edit file curdir/C_src/problem.c to define your own objective function.

  In particular, modify the subroutines

  setdim : which sets problem dimension setbounds : which sets upper and lower bounds on the variables funct : which defines the objective function

• PYTHON VERSION OF THE CODE:

  Edit file curdir/PYTHON/problem.py to define your own objective function.

  In particular, modify functions

  setbounds : which sets upper and lower bounds on the variables

  funct : which defines the objective function

2. At command prompt in curdir execute

   $> make

   which will create the executables 'acrs_c' (using C sources) and 'acrs_f' (using FORTRAN90 sources)

3. execute

   $> ./acrs_f

   $> ./acrs_c

   $> python PYTHON/main.py

4. MATLAB VERSION OF THE CODE

   In the folder curdir/Matlab there is a matlab version (acrs.m) of the code. File main.m contains an example usage of ACRS for Matlab to minimize the funcion defined in powell.m