# Optimizing minimization with fmincon function

2 views (last 30 days)
ektor on 25 May 2019
Dear all,
I have this function which I minimize:
g=randn(1000,1);
u=randn(1000,1);
y=randn(1000,1);
options = optimoptions('fmincon','Display','off');
f = @(x) sum( ( y-x(1)*g-u*x(2) ).^2 );
nonlcon = @unitdisk;
x = fmincon(f,[0.2 0.02],[],[],[],[],[],[],nonlcon,options);
where
function [c,ceq] = unitdisk(x)
c = - x(2) +0.01;
ceq = [];
Is there a faster way of doing this minimization?

Sulaymon Eshkabilov on 25 May 2019
Hi,
By setting up the solver algorithm in the option settings, the simulation time cna be shortened substantially. E.g.
g=randn(1000,1);
u=randn(1000,1);
y=randn(1000,1);
options = optimoptions('fmincon','Display','off', 'Algorithm', 'active-set');
f = @(x) sum( ( y-x(1)*g-u*x(2) ).^2 );
nonlcon = @unitdisk;
x = fmincon(f,[0.2 0.02],[],[],[],[],[],[],nonlcon,options);
This algorithm shortens the computation time by about 50%. If you are not satisfied with this, you can investigate furthermore with the option settings for fmincon.
Good luck.

ektor on 25 May 2019
Indeed it is much faster now.
What should be the 'options' settings for the 'fminsearch'?
Could you give me an 'options' example for this function?
Sulaymon Eshkabilov on 26 May 2019
hi,
Again if you post your example, that would be good to answer specifically w.r.t your problem constraints.