Derivative free optimization

This demo aims to demonstrate how certain derivative-free optimization methods work. The idea of these methods is to find the minimum of a function without relying on derivatives of the function.

Try modifying the parameters to see their effect, or hover/click them for an explanation. You can also try different methods on different functions to see their typical behavior. Certain methods may work better than others in practice, especially on more difficult problems. The chart displays a contour map of the optimization landscape. The optimal solution is within the most lines.

inertia

0.60

cognitive

0.50

social

0.50

convergence progress0%
|||||
||

about the method

The Nelder-Mead method is a simplex-based direct search method for nonlinear optimization. It is a numerical technique that does not require the computation of derivatives, making it suitable for optimizing functions where the gradients are not available or computationally expensive to calculate. In comparison to DE and PSO, it is not a global optimization method and it might get stuck to a local minimum.