# help using matlab solve ga function for optimization of non linear expression

9 views (last 30 days)
fima v on 4 Nov 2022
Edited: Matt J on 5 Nov 2022
Hello , i have this non linear function called "h" shown in the code bellow.
I want to optimize k1, k2 ,a ,b parameters so this function will get values between 0.55 and 0.45 over the x axes.
I have tried to use the following manual and save the script file as optim.m
i tried to use use solve ga to get my function be get values between 0.55 and 0.45 over the x axes
i get an error shown bellow.
Where did i go wrong?
"Error using optim
Constraints must be an OptimizationConstraint or a struct containing OptimizationConstraints."
Thanks.
h=@(a,b,k1,k2)abs(a*exp(-j*k1*x)+b*exp(-j*k2*x));
cam=h(a,b,k1,k2);
prob = optimproblem("Objective",cam);
prob.Constraints.cons1=a<=1;
prob.Constraints.cons2=a>=-1;
prob.Constraints.cons3=b<=1;
prob.Constraints.cons4=b>=-1;
prob.Constraints.cons5=h<=0.55;
prob.Constraints.cons5=h>=0.45;
[sol,fval] = solve(prob,"Solver","ga");

Matt J on 4 Nov 2022
Edited: Matt J on 4 Nov 2022
See, Optimize with Nonlinear Constraints Using GA. It should look something like this:
x=linspace(0,100,10000);
vars = ga(fun,4,A,b,Aeq,beq,lb,ub, @(vars)nonlcon(vars,x), options);
function [c,ceq]=nonlcon(vars,x)
a=vars(1); b=vars(2);
k1=vars(3); k2=vars(4);
h =abs(a*exp(-j*k1*x)+b*exp(-j*k2*x));
c(1)=0.45-min(h);
c(2)=max(h)-0.55;
ceq=[];
end
Matt J on 5 Nov 2022
Edited: Matt J on 5 Nov 2022
Optimization variables cannot be combined with complex data.
This ran to completion for me:
a=optimvar('a','Lower',-1,'Upper',+1);
b=optimvar('b','Lower',-1,'Upper',+1);
k1=optimvar('k1','LowerBound',-1,'UpperBound',1);
k2=optimvar('k2','LowerBound',-1,'UpperBound',1);
h=@(a,b,k1,k2)abs(a*exp(-j*k1*x)+b*exp(-j*k2*x));
hmax=@(a,b,k1,k2) max(h(a,b,k1,k2));
hmin=@(a,b,k1,k2) min(h(a,b,k1,k2));
cam=fcn2optimexpr( h, a,b,k1,k2);
camMax=fcn2optimexpr( hmax, a,b,k1,k2);
camMin=fcn2optimexpr( hmin, a,b,k1,k2);
prob = optimproblem("Objective",camMax);
prob.Constraints.camUpper=camMax<=0.55;
prob.Constraints.camLower=camMin>=0.45;
[sol,fval] = solve(prob,"Solver","ga");
Solving problem using ga. Optimization terminated: average change in the fitness value less than options.FunctionTolerance and constraint violation is less than options.ConstraintTolerance.
you have changed the original code.could you please use one "h" function?
No, you need multiple functions. However, here is a slightly different formulation which avoids hmin.
a=optimvar('a','Lower',-1,'Upper',+1);
b=optimvar('b','Lower',-1,'Upper',+1);
k1=optimvar('k1','LowerBound',-1,'UpperBound',1);
k2=optimvar('k2','LowerBound',-1,'UpperBound',1);
h=@(a,b,k1,k2)abs(a*exp(-j*k1*x)+b*exp(-j*k2*x));
hmax=@(a,b,k1,k2) max(h(a,b,k1,k2));
cam=fcn2optimexpr( h, a,b,k1,k2);
camMax=fcn2optimexpr( hmax, a,b,k1,k2);
prob = optimproblem("Objective",camMax);
prob.Constraints.camUpper=camMax<=0.55;
prob.Constraints.camLower=cam>=0.45;
[sol,fval] = solve(prob,"Solver","ga");