hexed::math::Objective_finite_diff Struct Reference

Estimates the derivatives of an optimization objective function by finite difference. More...

#include <math.hpp>

Classes
struct	Objective_sample

Public Member Functions
	Objective_finite_diff (std::function< Objective_sample(Mat<>)> objective_fun, Mat<> at, double finite_diff, double min_diff_ratio=1e-8)
	Estimates derivatives of objective_fun at the point at with a finite difference of finite_diff.

Public Attributes
bool	feasible
	is at a feasible point?
double	objective
	if feasible, then the objective at point at, otherwise unspecified
Mat	gradient
	gradient at point at
Mat< dyn, dyn >	hessian
	Hessian matrix at point at

Detailed Description

Estimates the derivatives of an optimization objective function by finite difference.

Constructor & Destructor Documentation

Objective_finite_diff()

hexed::math::Objective_finite_diff::Objective_finite_diff	(	std::function< Objective_sample(Mat<>)>	objective_fun,
		Mat<>	at,
		double	finite_diff,
		double	min_diff_ratio = 1e-8 )

Estimates derivatives of objective_fun at the point at with a finite difference of finite_diff.

objective_fun should take vectors of the same size as at and return an Objective_sample indicating the values of the objective function and also whether the argument is a feasible point. If Objective_sample::feasible is false, then Objective_sample::objective is irrelevant and may be set to any value you choose. All sample points p will satisfy std::abs(p(i) - at(i)) <= finite_diff for all valid indices i. If any of the sampled points are not feasible, the finite difference will be reduced and another attempt will be made. If the finite difference is reduced by a factor exceeding min_diff_ratio, it will give up and declare the point infeasible.

---

The documentation for this struct was generated from the following files:

• /home/micaiah/Desktop/hexed/include/math.hpp
• /home/micaiah/Desktop/hexed/libhexed/math.cpp