Provide partial gradient and Hessian to fminunc.

I have the gradient and Hessian in closed form only for a subset of the variables of my objective function. As an example consider

function [f, g, H] = rosenboth(x)

f = 100*(x(2) - x(1)^2)^2 + (1-x(1))^2;

g = -400*(x(2)-x(1)^2)*x(1)-2*(1-x(1));

H = 1200*x(1)^2-400*x(2)+2;

end

where g and H are only w.r.t. x(1). If g and H where the full gradient and Hessian, I would simply use fminunc using

options = optimoptions('fminunc','Algorithm','trust-region',...

'SpecifyObjectiveGradient',true,'HessianFcn','objective');

It it possible to provide fminunc only with a partial gradient and Hessian and let it do numerical gradient and Hessian estimation for the rest?

Answer 1

[Torsten]

It it possible to provide fminunc only with a partial gradient and Hessian and let it do numerical gradient and Hessian estimation for the rest?

No. But you could supply the rest by your own finite-difference approximation.