Matlab Fitness Function Genetic Algorithm Error

Hey everyone, I am trying to use genetic algorithm in matlab to perform SMM. However, I am having difficulties with passing extra parameters through the ga function. I have successfully used this function (and successfully passed the parameters through) with lsqnonlin, so I suspect there is some formatting error I am not understanding.

Here is where I call the function using ga:

xL=[ 0, 0, 0, 0, 0, 0,   0, 0, 0, 0,  0,  0,0,0,0,0,0,0,1.5,1.5];
xU=[.5,.5,.6,.6,.95,10,5, 5, .15, .15,.95,1,1,.95,.95,2,2,10,7,7];
FitFunc=@(x) lsqnonlindiscretecalibrate_V2_AR(x,A1,A2,r,discount,fracc,matchfrac1,matchfrac2,wageprem,N,targetu1,targetu2,sepfrac,thetatarget,...
  percent1,percent2,moverate_target,jfptarget,jsptarget,lstarget,avgutarget,replacetarget,etarget,duration_target1,duration_target2,...
  fsize,emplyfrac_target,lag1target,lag2target,ssq1target,ssq2target,durtarget,sims,burn,yfrac_target);
options2=optimoptions(@ga,'Display','Iter','FunctionTolerance',1e-30,'MaxGenerations',3,'UseParallel',true);
[vars_parameters,error_param]=ga(FitFunc,20,[],[],[],[],[],xL,xU,options2);

Here is the beginning of the function file itself:

 function SSRc=lsqnonlindiscretecalibrate_V2_AR(x,A1,A2,r,discount,fracc,matchfrac1,matchfrac2,wageprem,N,targetu1,targetu2,sepfrac,thetatarget,percent1,percent2,moverate_target,jfptarget,jsptarget,lstarget,avgutarget,replacetarget,etarget,duration_target1,duration_target2,fsize,emplyfrac_target,lag1target,lag2target,ssq1target,ssq2target,durtarget,sims,burn,yfrac_target)
z1=x(1);
z2=x(2);
rho1=x(3);
rho2=x(4);
beta=x(5);
k1=x(6);
mu1=x(7);
mu2=x(8);
deltac=x(9);
deltad=x(10);
ps=x(11);
alpha1=x(12);
alpha2=x(13);
Lp1=x(14);
Lp2=x(15);
psig1=x(16);
psig2=x(17);
k2=x(18);
p1bar=x(19);
p2bar=x(20);
....Some stuff...  end

And here is the error I get:

Error using functionHandleOrCell (line 12) The constraint function must be a function handle. Error in validate (line 228) [nonlcon,NonconFcnArgs] = functionHandleOrCell('NonconFcn',nonlcon);

Error in gacommon (line 65) [options,nvars,FitnessFcn,NonconFcn] = validate(options,type,nvars,fun,nonlcon,user_options); Error in ga (line 336) NonconFcn,options,Iterate,type] = gacommon(nvars,fun,Aineq,bineq,Aeq,beq,lb,ub, ... Error in discrete_calibration_V2_AR (line 92) [vars_parameters,error_param]=ga(FitFunc,20,[],[],[],[],[],xL,xU,options2);

Apologies for the lengthy names, I need to keep everything straight. I understand that I am inputting something in a way Matlab doesn't like, but I cannot figure out what it should be. I have read the documentation for passing parameters in ga, and I think this is correct. Any help would be greatly appreciated, I assume it's something simple I do not understand/missed!

Answer 1

[Alan Weiss]

Your ga call is not correct:

ga(FitFunc,20,[],[],[],[],[],xL,xU,options2);

It should be

ga(FitFunc,20,[],[],[],[],xL,xU,[],options2);

Sometimes it is easier to keep track of things by adding empty variables:

A = [];
b = [];
Aeq = [];
beq = [];
nonlcon = [];
[vars_parameters,error_param]=ga(FitFunc,20,A,b,Aeq,beq,xL,xU,nonlcon,options2)

Alan Weiss

MATLAB mathematical toolbox documentation

[Ben Raymond]

I knew it was something small, thanks!

[Joseph Stalin]

Hey I was trying to do something similar except I am using a function handle which takes in an input vector x with all elements being binary(0 or 1). It is a quadratic mixed integer programmig problem however I cannot get it to work with the ga solver. Here is my function call:

fun = @(x) (x'*Co*x)-S'*x;
intcon = [];
for i=1:n
    intcon = [intcon i];
end

A = []; b = []; Aeq = []; beq = []; nonlcon = [];
lb = zeros(n,1);
ub = ones(n,1);
options = optimoptions('ga','PlotFcn',@gaplotbestf);
x = ga(fun,nres,A,b,Aeq,beq,lb,ub,nonlcon,intcon,options);

Here the input x is a nx1 vector, Co is nxn matrix and S is a nx1 vector. The output of the function fun is a scalar. Coud anyone point to anything I might have missed? is there any special syntax for vector arguments?