both real and integer variables ga optimization

How can i perform an optimization with the gamultiobj function using both real and integer variables ?

Answer 1

[Shashank Prasanna]

Andrea, integers are real, however I think you are referring to solving a mixed integer problem. Here is an example of how it can be done using GA:

http://www.mathworks.com/help/gads/mixed-integer-optimization.html

[andrea84]

Hi Benji, you're right I meant that some of my variables should be discrete.

Anyway if the variable x1 has lb=0 and ub=100 how can i constrain it so that it assume only some numerical value ?

e.g [0,20,40,60,80,100]

[andrea84]

I also tried the example written in the link you provided but it failed, giving this error:

??? Error using ==> ga
Too many input arguments.

I'm using the R2011a version, is this the problem?

[Walter Roberson]

R2011a did not support constraining to integer.

I don't think you can (even with R2012b) constrain to arbitrary sets of values. However, you can (in versions that support it) constrain to integer values, with a range of 1 to the number of different values, and then in the function you provide, use that variable as an index into the table of actual values.

[andrea84]

Thanks for the explanation!

I'm thinking to use a function to "map" the values obtained from the ga, e.g. [1 2 3 4 5] to the range i need: [20 40 60 80 100].

Answer 2

[Alan Weiss]

As Benji said, you cannot perform mixed integer optimization automatically in GA unless you have MATLAB R2011b or later. However, you can write custom population functions for this purpose; see this explanation. You might also want to see this example on mapping integer variables to a finite set of choices.

Good luck,

Alan Weiss

MATLAB mathematical toolbox documentation