Scaling Scheme.

Kian H Low uceckil at
Tue May 14 07:25:08 EDT 2002


I am applying Ga to solve design problem in chemical engineering.

Has anyone had any experience regarding the use of the sigma truncation
scaling scheme? As you know, the default value for the sigma truncation
multiplier in GAlib is 2. Has anyone experience using another value for
this and if so, what is the effect on the optimization, in terms of
overcoming premature convergence and slow finishing? I know the value
can be problem specific, but is there any rule of thumb? Will a
higher/lower value of the multiplier help, or is there a trade-off?

Also, for negative objective functions, this method is said to 'truncate
arbitrarily at 0', can someone please explain what this mean?

Thanks in advance.

Kian H Low
-------------- next part --------------
An HTML attachment was scrubbed...

More information about the galib mailing list