BLMVM


Software for Bound-Constrained Optimization

BLMVM minimizes an objective function subject to lower and upper bounds on the variables. Applications typically involve minimizing cost, energy, or error subject to legal, physical, or logistical constraints. Previous success stories have involved parameter estimation, molecular geometry, and risk management. Its convergence requires a continuous objective function and first derivatives over the feasible domain.

This implementation is written in C, although examples of its use in C, Fortran, Matlab, and Python are included in the distribution. BLMVM is suitable for single-processor and parallel architectures.

Download:

The current release BLMVM (Version 1.1) can be obtained by entering your email address below and downloading the compressed tar file that contains the software.

Email: .

Check the documentation for the latest installation instructions.

Reference:

@TechReport{blmvm,
       author = "Steven J. Benson and Jorge Mor\'{e}",
       title = "A Limited-Memory Variable-Metric Algorithm for Bound-Constrained Minimization",
       institution = "Mathematics and Computer Science Division, Argonne National Laboratory",
       number = "ANL/MCS-P909-0901",
       year = "2001",
}
PostScript.

The solver can also be used through NEOS, and TAO.

The numerical algorithm, its implementation, and examples in BLMVM were developed by Steven Benson, Jorge More', Jim Adduci, Matthew Beauregard.

BLMVM was funded through the MICS Division of the U.S. Department of Energy Office of Science.