Curve fitting by using of Minimization. Base class is CurveFitting, but it is virtual class, which you cannot use itself. You can use some derived classes already prepared, Line, Polynomial, Catenary, Exponential and InvCycloid class. Or you can create another derived class from "CurveFitting" class.

The usage is shown in "CurveFittingMai.cpp". First, call constructure, whose arguments are sample points cordinates, std::vector<double> x and std::vector<double> y. Next, prepare std::vector<double> as initial values of curve equation parameter. Eigen::VectorXd is also acceptable. Last, call solve() function. Its argument will be updated to the solution.

In the case of creating a new derived class, you should define its model() function. Because a template class cannot become a vitrual class, you should define 2 types of model() function. One is for AutomaticDifferentiation::FuncPtr<double>, the other is for double. Their return will be the curve equation, to which you want to fit the sample data. Additionally, you need to define num_param(), which is helpful to inform how many dimention the parameter vector has.