[Scilab-users] derivative vs numderivative

[paul.carrico at free.fr]

Hi all

I've not been using Scilab for optimization tasks for years and I
decided to dig up my old codes; I didn't follow such items in the
mailing list so the current questionning has probably ever been treated
(?!?!):

why numderivative leads to different result than derivative (since
derivative will be removed in the next release)? ... algorithms + step
calculation  are probably different (while they are using finite
difference method ... I don't know)

To myself first, I've made a (very) basic example based on the famous
Rosembrock function to perform tests 

Thanks for any support 

Paul

###############################################################################"
mode(0)
clear

// function
function f = rosembrock(x)
    f = ( 1 - x(1))^2 + 100*( x(2)-x(1)^2 )^2;
endfunction

// cost function
function [f,g,ind] = cost(x,ind)
    f = rosembrock(x);
//    g = derivative(rosembrock, x.',order = 4); 
    g = numderivative(rosembrock, x.',order = 4); 
endfunction

//initial_parameters = [10 10]
initial_parameters = [100 100]
//lower_bounds = [90 90];
lower_bounds = [0 0];
upper_bounds = [1000 1000];
//[fopt , xopt] = optim(cost,initial_parameters) 
[fopt , xopt] =
optim(cost,'b',lower_bounds,upper_bounds,initial_parameters,'ar',100,100)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/users/attachments/20170110/c56648c2/attachment.htm>