How to control the step length or accuracy of decision variables X in gamultiobj? I need the X just be assigned as 0.001, 0.002 or 0.02, instead of 0.001234.

I use the funtion gamultiobj. I need the decision variables X just be assigned as 0.001, 0.002 or 0.02, instead of 0.001234. Is there a method to control the step length or accuracy of decision variables X in the multiobjective optimization?

[X,Fval,exitFlag,Output]=gamultiobj(FitnessFunction,numberOfVariables,A,b,Aeq,beq,lb,ub,options);

Answer 1

[Walter Roberson]

gamultiobj is not designed for discrete variables.

What you have to do is use the options structure:

• pass in an CreationFcn that uses only the permitted values
• supply your own CrossoverFcn and MutationFcn, which "just happen" to satisfy your requirement each time

[Yuan Chen]

Thank you for your answer. Inspired by your suggestion, I add a nonlinear constraint:

function [c,ceq]=mycon(x)
c=[];
ceq=x-round(x,3);

[Sonja Kallio]

Could you give us an example how to write CreationFcn with prmitted values?

I have CreationFcn which uses a for loop to create the Initial Population.

There is 3 decision variables which are varied by the for loop.

Should I define the permitted values somehow in the function, for example like this:

x(1) = (1:1:10)

Best regards,

Sonja Kallio

[Walter Roberson]

Assuming for a moment that you are using integer variables and that lb and ub have been set appropriately,

for iter = 1 : population size
    for var = 1 : nvars
        x(var) = randi([lb(var), ub(var)]);
    end
    population(iter,:) = x;
end