Unrecognized variable, want to create a list of variable

Hello,

I'm trying to run an optimization model but I always get the error "Unrecognized function or variable 'c1_std'.

I want to call d = (c1_std, c2_std, c3_std) to run my optimization model and find the best parameters. How should I define d to call d(1)=c1_std in my different functions?

I'm not an expert in Matlab so please excuse me if my code is quite rude.

Thank you in advance.

Paul

muL = 2000;
sigL = 200;
R1 = 1-9.92*10^-5;
R2 = 1-1.2696*10^-4;
R3 = 1-3.87*10^-6;
d = [c1_std,c2_std,c3_std];

Sr1_min = sqrt(((((1.5-1)*muL)/norminv(R1))^2)-(sigL)^2);
Sr1_max = sqrt(((((2.5-1)*muL)/norminv(R1))^2)-(sigL)^2);
Sr2_min = sqrt(((((1.5-1)*muL)/norminv(R2))^2)-(sigL)^2);
Sr2_max = sqrt(((((2.5-1)*muL)/norminv(R2))^2)-(sigL)^2);
Sr3_min = sqrt(((((1.5-1)*muL)/norminv(R3))^2)-(sigL)^2);
Sr3_max = sqrt(((((2.5-1)*muL)/norminv(R3))^2)-(sigL)^2);

sqrt((((1.5-1)*muL)/norminv(R1)^2)-(sigL)^2)
norminv(R1)

lb = [Sr1_min,Sr2_min,Sr3_min];
ub = [Sr1_max,Sr2_max,Sr3_max];
A = [];
B = [];
Aeq = [];
Beq = [];
d0 = (lb+ub)/2;
fun = @(d) parameterfun(d,muL,sigL);
const = @(d) nonlcon(d,muL,sigL);
[d,fval] = fmincon(fun,d0,A,B,Aeq,Beq,lb,ub,const);

function Rs = parameterfun(d,muL,sigL)

muL = 2000;
sigL = 200;
R1 = 1-9.92*10^-5;
R2 = 1-1.2696*10^-4;
R3 = 1-3.87*10^-6;
d = [c1_std,c2_std,c3_std];

mu_Sr1 = muL+norminv(R1)*sqrt((sigL)^2+(d(1))^2);
mu_Sr2 = muL+norminv(R2)*sqrt((sigL)^2+(d(2))^2);
mu_Sr3 = muL+norminv(R3)*sqrt((sigL)^2+(d(3))^2);

Y1_mean = muL-mu_Sr1;
Y2_mean = muL-mu_Sr2;
Y3_mean = muL-mu_Sr3;

Y1_std = sqrt((d(1))^2+(sigL)^2);
Y2_std = sqrt((d(2))^2+(sigL)^2);
Y3_std = sqrt((d(3))^2+(sigL)^2);

Y_mean = [Y1_mean Y2_mean Y3_mean];
Y_std = [(Y1_std^2) (sigL)^2 (sigL)^2; (sigL)^2 (Y2_std)^2 (sigL)^2; (sigL)^2 (sigL)^2 (Y3_std)^2];

y = zeros(size(d));
Rs = mvncdf(y, Y_mean, Y_std);

end

function [c,ceq] = nonlcon(d,muL,sigL)

muL = 2000;
sigL = 200;
R1 = 1-9.92*10^-5;
R2 = 1-1.2696*10^-4;
R3 = 1-3.87*10^-6;
d = [c1_std,c2_std,c3_std];

c(1) = ns_min(1) - ((muL+norminv(R1)*sqrt((d(1)^2)+(L_std^2)))/muL);
c(2) = ns_min(2) - ((muL+norminv(R2)*sqrt((d(2)^2)+(L_std^2)))/muL);
c(3) = ns_min(3) - ((muL+norminv(R3)*sqrt((d(3)^2)+(L_std^2)))/muL);
c(4) = ((muL+norminv(R1)*sqrt((d(1)^2)+(L_std^2)))/muL) - ns_max(1);
c(5) = ((muL+norminv(R2)*sqrt((d(2)^2)+(L_std^2)))/muL) - ns_max(2);
c(6) = ((muL+norminv(R3)*sqrt((d(3)^2)+(L_std^2)))/muL) - ns_max(3);
c(7) = c_min(1) - (c1_std/((muL+norminv(R1)*sqrt((d(1)^2)+(L_std^2)))));
c(8) = c_min(2) - (c2_std/((muL+norminv(R2)*sqrt((d(2)^2)+(L_std^2)))));
c(9) = c_min(3) - (c3_std/((muL+norminv(R3)*sqrt((d(3)^2)+(L_std^2)))));
c(10) = (c1_std/((muL+norminv(R1)*sqrt((d(1)^2)+(L_std^2))))) - c_max(1);
c(11) = (c2_std/((muL+norminv(R2)*sqrt((d(2)^2)+(L_std^2))))) - c_max(2);
c(12) = (c3_std/((muL+norminv(R3)*sqrt((d(3)^2)+(L_std^2))))) - c_max(3);

format long
c

ceq = [];

end

[Jan]

In the line:

d = [c1_std,c2_std,c3_std];

neither Matlab nor the readers in the forum can guess, what c1_std, c2_std, c3_std is. What are these variables or functions? Where are they defined?

[Paul AGAMENNONE]

This is what the optimization model is supposed to search when it optimizes Rs.

Thus I don't know how to call them in my example

Answer 1

[Stephen23]

The basic problem appears to be your attempt to re-define these parameters again inside every function, thus obliterating the input values. The simple solution is to define them only once and then pass them as input arguments:

https://www.mathworks.com/help/optim/ug/passing-extra-parameters.html

The d value seems to be what you are trying to optimize, so you should not specify any values for that, because the d values are provided by FMINCON.

muL = 2000;
sigL = 200;
R1 = 1-9.92*10^-5;
R2 = 1-1.2696*10^-4;
R3 = 1-3.87*10^-6;

Sr1_min = sqrt(((((1.5-1)*muL)/norminv(R1))^2)-(sigL)^2);
Sr1_max = sqrt(((((2.5-1)*muL)/norminv(R1))^2)-(sigL)^2);
Sr2_min = sqrt(((((1.5-1)*muL)/norminv(R2))^2)-(sigL)^2);
Sr2_max = sqrt(((((2.5-1)*muL)/norminv(R2))^2)-(sigL)^2);
Sr3_min = sqrt(((((1.5-1)*muL)/norminv(R3))^2)-(sigL)^2);
Sr3_max = sqrt(((((2.5-1)*muL)/norminv(R3))^2)-(sigL)^2);

lb = [Sr1_min,Sr2_min,Sr3_min];
ub = [Sr1_max,Sr2_max,Sr3_max];
A = [];
B = [];
Aeq = [];
Beq = [];
d0 = (lb+ub)/2;
fun = @(d) parameterfun(d,muL,sigL,R1,R2,R3);
const = @(d)    nonlcon(d,muL,sigL,R1,R2,R3);
[d,fval] = fmincon(fun,d0,A,B,Aeq,Beq,lb,ub,const)

Converged to an infeasible point.

fmincon stopped because the size of the current step is less than
the value of the step size tolerance but constraints are not
satisfied to within the value of the constraint tolerance.


Consider enabling the interior point method feasibility mode.

d = 1×3

179.9441 186.3380 103.6089

fval = 0.9998

That works without error. Only you can check if it makes sense.

function Rs = parameterfun(d,muL,sigL,R1,R2,R3)
%
mu_Sr1 = muL+norminv(R1)*sqrt((sigL)^2+(d(1))^2);
mu_Sr2 = muL+norminv(R2)*sqrt((sigL)^2+(d(2))^2);
mu_Sr3 = muL+norminv(R3)*sqrt((sigL)^2+(d(3))^2);
%
Y1_mean = muL-mu_Sr1;
Y2_mean = muL-mu_Sr2;
Y3_mean = muL-mu_Sr3;
%
Y1_std = sqrt((d(1))^2+(sigL)^2);
Y2_std = sqrt((d(2))^2+(sigL)^2);
Y3_std = sqrt((d(3))^2+(sigL)^2);
%
Y_mean = [Y1_mean Y2_mean Y3_mean];
Y_std = [(Y1_std^2) (sigL)^2 (sigL)^2; (sigL)^2 (Y2_std)^2 (sigL)^2; (sigL)^2 (sigL)^2 (Y3_std)^2];
%
y = zeros(size(d));
Rs = mvncdf(y, Y_mean, Y_std);
%
end

function [c,ceq] = nonlcon(d,muL,sigL,R1,R2,R3)
%
% You did not define these, so I will presume all have value 1:
L_std = 1;
ns_min = ones(1,3);
ns_max = ones(1,3);
c_min = ones(1,3);
c_max = ones(1,3);
c1_std = 1;
c2_std = 1;
c3_std = 1;
%
c(1) = ns_min(1) - ((muL+norminv(R1)*sqrt((d(1)^2)+(L_std^2)))/muL);
c(2) = ns_min(2) - ((muL+norminv(R2)*sqrt((d(2)^2)+(L_std^2)))/muL);
c(3) = ns_min(3) - ((muL+norminv(R3)*sqrt((d(3)^2)+(L_std^2)))/muL);
c(4) = ((muL+norminv(R1)*sqrt((d(1)^2)+(L_std^2)))/muL) - ns_max(1);
c(5) = ((muL+norminv(R2)*sqrt((d(2)^2)+(L_std^2)))/muL) - ns_max(2);
c(6) = ((muL+norminv(R3)*sqrt((d(3)^2)+(L_std^2)))/muL) - ns_max(3);
c(7) = c_min(1) - (c1_std/((muL+norminv(R1)*sqrt((d(1)^2)+(L_std^2)))));
c(8) = c_min(2) - (c2_std/((muL+norminv(R2)*sqrt((d(2)^2)+(L_std^2)))));
c(9) = c_min(3) - (c3_std/((muL+norminv(R3)*sqrt((d(3)^2)+(L_std^2)))));
c(10) = (c1_std/((muL+norminv(R1)*sqrt((d(1)^2)+(L_std^2))))) - c_max(1);
c(11) = (c2_std/((muL+norminv(R2)*sqrt((d(2)^2)+(L_std^2))))) - c_max(2);
c(12) = (c3_std/((muL+norminv(R3)*sqrt((d(3)^2)+(L_std^2))))) - c_max(3);
%
ceq = [];
%
end

[Paul AGAMENNONE]

Thank you Stephen for your answer, I finally understand and the program works perfectly :)