Gradient
The classes ForwardGradient, BackwardGradient and SymetricGradient allow to approximate the value of the gradient of a differentiable function.
// Create the function
const auto f = [](const Vect<double> &x) {
return sin(x(0));
};
// Create its gradient
const auto grad = ForwardGradient<decltype(f)>(f, 1e-7);
// Compute the value of the gradient at a given point
const double value = grad(std::vector<double>{1})(0);
-
template<typename F>
class ForwardGradient - #include <gradient.hpp>
Functor computing the gradient of a function with a given value of epsilon using the method of forward gradient. The function f is evaluated (n + 1) times, where n is the dimension of the space.
Public Functions
-
template<typename F>
class BackwardGradient - #include <gradient.hpp>
Functor computing the gradient of a function with a given value of epsilon using the method of backward gradient. The function f is evaluated (n + 1) times, where n is the dimension of the space.
Public Functions
-
template<typename F>
class SymetricGradient - #include <gradient.hpp>
Functor computing the gradient of a function with a given value of epsilon using the method of symetric gradient. The function f is evaluated 2n times, where n is the dimension of the space. However, the method is more precise than the others.
Public Functions