Error using MATLAB GA - [left and right sides have a different number of elements]

I'm running the MATLAB GA to optimise my objective function using:

var1 = (linspace(100, 400, 10));
var2 = (linspace(2, 7, 10));

for i=1:length(var2)
    for j=1:length(var1)
        
        objFun = @(x) my_objFun(var1(j),var2(i));
        constraintFun = @(x) my_constraintFun(var1(j),var2(i));
        
        try
            [x,fval,exitflag,output]= ga(objFun,nvars,A,b,Aeq,beq,lb,ub,constraintFun,IntCon,options);
        catch e
            fprintf(1,'[ERROR] - The identifier was:\n%s\n',e.identifier);
            fprintf(1,'[ERROR] - There was an error! The message was:\n%s\n',e.message);
        end
    end
end

The GA function is called repetitive times inside a loop. It sometimes runs successfully, other times it throws ERROR 1 and in a few other occasions it throws ERROR 2.

• ERROR 1

[ERROR] - The identifier was:
MATLAB:matrix:singleSubscriptNumelMismatch
[ERROR] - There was an error! The message was:
Unable to perform assignment because the left and right sides have a different number of elements.

• ERROR 2

[ERROR] - The identifier was:
MATLAB:griddedInterpolant:NonMonotonicCompVecsErrId
[ERROR] - There was an error! The message was:
Sample points must be unique and sorted in ascending order.

Any idea what that could be?

I'm using try and catches all over my my_objFun and my_constraintFun to make sure this error is not coming from there.

[Raymond Norris]

My guess is that it's not in your objective or constraint function, but what is getting returned from ga. I'd try two things:

1. First try calling your for-loop in reverse order to see if that reproduces it. If so, it'll be much easier to debug.
2. Next, call ga, but different output values. For instance

[x] = ga();
[~,fval] = ga();
[~,~,exitflag] = ga();
[~,~,~,output] = ga();

My sense is that the last example may reproduce the error. You could try this with both a reverse order for-loop and a parfor.

[Matheus Torquato]

Thanks. I'll try that out. Not sure how calling the for-loop in reverse order may help though.

[Matheus Torquato]

@Raymond Norris

Running it again using the for loop in reverse order did not change anything.

I did not see any difference as well when calling the GA with different outputs.

var1 = (linspace(100, 400, 10));
var2 = (linspace(2, 7, 10));
for i=length(var2):-1:1
    for j=length(var1):-1:1
        objFun = @(x) my_objFun(x,var1(j),var2(i));
        constraintFun = @(x) my_constraintFun(x,var1(j),var2(i));
        try
            [~,fval,~,output]= ga(objFun,nvars,A,b,Aeq,beq,lb,ub,constraintFun,IntCon,options);
        catch e
            fprintf(1,'[ERROR] - The identifier was:\n%s\n',e.identifier);
            fprintf(1,'[ERROR] - There was an error! The message was:\n%s\n',e.message);
        end
    end
end

[Alvaro]

Could you remove the try-catch block and post the full error stack?

Answer 1

[Alan Weiss]

I do not understand this line:

objFun = @(x) my_objFun(var1(j),var2(i));

Where does the variable x enter objFun? I think that is your main problem. Try calling

objFun(x0)

for some x0 that is a row vector with nvars elements. It probably will not return what you hope it returns.

Alan Weiss

MATLAB mathematical toolbox documentation

[Matheus Torquato]

Hi @Alan Weiss,

Thank you for your comment.

I'm sorry about that. That was a typo while writing this snippet pseudocode.

I've corrected

        objFun = @(x) my_objFun(var1(j),var2(i));
        constraintFun = @(x) my_constraintFun(var1(j),var2(i));

to

        objFun = @(x) my_objFun(x,var1(j),var2(i));
        constraintFun = @(x) my_constraintFun(x,var1(j),var2(i));

This latter one is what I have been using in my code.