Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.


Dave
Posts:
2
Registered:
2/12/14


Genetic Algorithms  Mixing doubles and integers
Posted:
Feb 12, 2014 7:51 PM


Hi,
quoting from: http://www.mathworks.co.uk/help/gads/mixedintegeroptimization.html#bs1cifg
under "Characteristics of the Integer ga Solver":
"ga uses only the binary tournament selection function (SelectionFcn option), and overrides any other setting."
Is there a theoretical restriction why other tournament sizes can't be used? e.g., ternary ones? or is this just the way the implementation was chosen to be?
I tried following ga.m through but eventually got stuck at gaminpl.p , so I couldn't figure out why the prevention from using other tournament sizes for the selection.
If there is a way to get around it (assuming no theoretical restriction) I'd be grateful to hear of it (supposedly, one can always use a double instead of an int and have the fitness function round to an int but that would be suboptimal).
Thanks in advance. Dave.



