nelder-mead

Cpp implementation of the Nelder-Mead optimization algorithm. Inspired by : https://github.com/fchollet/nelder-mead. That's make you able to try it very quickly on your cpp project which use Eigen as matrix computing library.

Example.cpp

First, you have to define your loss function that take an Eigen::VectorX as input. You can extract all feature you want from it with direct access. Here it's an example with a distance calculation.

double function(Eigen::Matrix<double, 3, 1> x){
    Vector<3> target(2,1,3);
    Vector<3> dist = target-x;
    return dist.dot(dist);
}

After, define your start point according to your input dimenssion.

Eigen::Matrix<double, 3, 1> start(0.0,0.0,0.0);

Then just call the Nelder-Mead function with the right template argument corresponding to your input dimension. You can check at Wikipedia page to understand all constants influances.

auto res = Nelder_Mead_Optimizer<3>(function, start, 0.1, 10e-10);

Reference

See the description of the Nelder-Mead algorithm on Wikipedia: https://en.wikipedia.org/wiki/Nelder%E2%80%93Mead_method

Enderdead/nelder-mead

nelder-mead

Example.cpp

Reference