Public Domain Computer Codes for Optimization

I have downloaded the following public domain codes for solving optimization problems.

Minpack routines

Minpack routines are Fortran codes for Levenberg-Marquardt for nonlinear least-squares problems. They differ in terms of gradient and Jacobian evaluation and number of user-provided parameters.

Comments on Minpack routines:

(1) Before using, check out the settings in the DPMPAR function; uncomment those statements which apply to your machine.

(2) These routines make far too many tests of equality between a real number (about to be used as a divisor) and zero. These should be replaced by tests of whether a number is less than machine epsilon; otherwise overflow is a danger.

LMDIF Allows the user to provide only the objective and then computes the gradient and Jacobian by finite-difference formulas.

LMDIF1 A simpler driver for HYBRD

LMDER Assumes that the user provides gradient and Jacobian routines as well as the objective.

LMDER1 A simpler driver for HYBRDJ

Matlab BFGS Code

Chris Sims offers matlab code for implementing BFGS.

Tensor Method Minimization

TENMIN generalizes Newton's method to include some nonquadratic terms. The result is code which is better behaved on problems which have nearly singular Hessians near the solution.

Constrained Optimization Code

Box constrained optimization code. This file contains a TeX paper and Fortran code.

Sparse Tensor Method Minimization