Genetic programming doubts

Francesco Patanè

unread,

Sep 7, 2021, 3:30:40 PM9/7/21

to HeuristicLab

Hi everyone,

I am using HL to obtain an equation like f=f(x1,x2,...). In my case one variable has a small correlation, therefore the algorithm doesn't use it in the equation. Since it is important that the variable appears in my equation, is there a way to force the algorithm to use it?

In addition, I'd like to get deep into the methods used. I want to know how each analyzer, cross operator, ecc. work. How can I get more information about this?

Thanks for answers in advance.

Gabriel Kronberger

unread,

Sep 9, 2021, 11:28:17 AM9/9/21

to noreply-spamdigest via HeuristicLab

Hello,

the capability of GP for automatic selection of relevant features is considered an advantage.

There are ways to force that a feature occurs in GP model by fixing the structure. This can be enforce in HL via the grammar but it is not easy in the GUI.

I would rather recommend that you extend the runtime of your algorithm (increase population size, increase number of generations). If the feature is actually relevant GP will be able to pick it up if you let it run for long enough even if the benefit of the feature is small.

Best, Gabriel

--
You received this message because you are subscribed to the Google Groups "HeuristicLab" group.
To unsubscribe from this group and stop receiving emails from it, send an email to heuristiclab...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/heuristiclab/7470fa5f-9072-4b16-a4d1-f6baa66a4494n%40googlegroups.com.

Francesco Patanè

unread,

Sep 12, 2021, 12:20:55 PM9/12/21

to HeuristicLab

Hi Gabriel,

thanks you very much for the answer. I know that it is an advantage but, just in this case I do not need it. I will try as you told me.

About the theory of the analyzers, operators and mutators, can't you suggest me anything? I'm writing a thesis and I'd like to give some theory notions but the Dumitrescu book reported as bibliography is not easily accessible for me.

Any suggestion is welcome.

Kind regards,

Francesco

Gabriel Kronberger

unread,

Sep 13, 2021, 6:02:35 AM9/13/21

to noreply-spamdigest via HeuristicLab

Hi,

I can recommend the Field Guide to Genetic Programming. https://books.google.at/books?id=3PBrqNK5fFQC&lpg=PP1&dq=field%20guide%20to%20genetic%20programming&hl=de&pg=PP1#v=onepage&q=field%20guide%20to%20genetic%20programming&f=false

Many of the algorithmic details in HeuristicLab are described in our book "Genetic Algorithms and Genetic Programming" https://books.google.at/books?id=EkELtZAXViEC&lpg=PP1&ots=rtVN7uTzQP&dq=genetic%20algorithms%20and%20genetic%20programming&lr&hl=de&pg=PP1#v=onepage&q=genetic%20algorithms%20and%20genetic%20programming&f=false

Best, Gabriel

To view this discussion on the web visit https://groups.google.com/d/msgid/heuristiclab/e0a71371-c195-451d-a569-5772909cc5e2n%40googlegroups.com.

Francesco Patanè

unread,

Sep 16, 2021, 6:54:28 AM9/16/21

to HeuristicLab

Thank you very much Gabriel.

Kind regards,

Francesco

Reply all

Reply to author

Forward