Levmar

Levmar is a cool software package that implements the Levenberg-Marquardt nonlinear least squares algorithms in C and C++. It's pretty handy for optimization tasks!

History of Levmar

The lmder routine from Minpack, developed back in the early '80s at Argonne National Lab, is one of the most popular free versions of the LM algorithm. It was originally written in FORTRAN77 and has stood the test of time as reliable software.

Why Use C/C++?

You might wonder why we need a version in C when we already have FORTRAN. Well, calling FORTRAN from C can get complicated with rules about name mangling, argument passing, and memory layouts. Plus, not every programming environment has a FORTRAN compiler available. Sometimes, understanding FORTRAN code can be tricky too—especially for those who aren't familiar with it. Even automatic translators don’t always make it clear.

The Benefits of Levmar

This is where Levmar comes to save the day! It includes various flavors of LM implementations that work great with both C and C++. You get double and single precision options, plus analytic and finite difference approximated Jacobians—all for free under the GNU General Public License!

User-Friendly Functions

Levmar provides easy-to-use functions following a simple naming rule: The first letter (d or s) shows if it's double or single precision, while the suffix (_der or _dif) tells you if it's using an analytic or approximate Jacobian.

Optimization Options

• dlevmar_der(): double precision with analytic Jacobian


• dlevmar_dif(): double precision with finite difference approximated Jacobian


• slevmar_der(): single precision with analytic Jacobian


• slevmar_dif(): single precision with finite difference approximated Jacobian

Constrained Optimization Features

If you're dealing with constraints, Levmar has you covered too! Check these out:

• dlevmar_lec_der(): double precision linear equation constraints with analytic Jacobian


• slevmar_bc_der(): single precision box constraints with analytic Jacobian


• dlevmar_blec_dif(): double precision box & linear equation constraints using finite differences


• slevmar_blec_der(): single precision box & linear equation constraints using analytic derivatives