cmss

Mini-HOWTO on using Octave for Unconstrained Nonlinear Optimization1

Nonlinear optimization problems are very common and when a solution cannot be found analytically, one usually tries to find it numerically. This document shows how to perform unconstrained nonlinear minimization using the Octave language for numerical computation. We assume to be so lucky as to have an initial guess from which to start an iterative method, and so impatient as to avoid as much as possible going into the details of the algorithm. In the following examples, we consider multivariable problems, but the single variable case is solved in exactly the same way.

All the algorithms used below return numerical approximations of local minima of the optimized function. In the following examples, we minimize a function with a single minimum (Figure 1), which is relatively easily found. In practice, success of optimization algorithms greatly depend on the optimized function and on the starting point.





Søren Hauberg 2008-04-29