For my diploma thesis I needed an easy-to-use optimisation algorithm which could minimise a given function. I had access to Matlab, but surprisingly, none of the supplied optimisation functions seemed to satisfy my needs: I wanted to globally minimise a non-linear function within boundaries without using gradients.
So I did a little research and found a Globalized Nelder-Mead method, which is a tuned version of the good old downhill simplex algorithm. Then I implemented it as a Matlab function. By the way: the function also works fine with Octave.

