Print

Print


Hi Esmail,
if you want to compare your algorithm on some continuous test functions, I
suggest the benchmark suite introduced at the CEC 2005 competition on
numerical optimization. The suite includes 25 well-chosen test functions.
The source code of them in several programming languages is provided at
http://www3.ntu.edu.sg/home/EPNSugan/index_files/CEC-05/CEC05.htm (look for
the right link).

The specification of the suite includes everything that is relevant:
Complete definition of the functions including the initialization of your
EA, the search interval, dimensionality, termination criteria (this answers
your previous post), even the statistical output and its formatting. If you
stick to these guidelines you can compare your method easily to many other
algorithms.

Here is a link with the specification:
http://www.bionik.tu-berlin.de/user/niko/Tech-Report-May-30-05.pdf

There are also a couple of more competitions recently. Have look at e.g.
http://www3.ntu.edu.sg/home/EPNSugan/ to find some more.

Best, Stefan







On Sun, Jan 17, 2010 at 1:44 PM, Esmail Bonakdarian <[log in to unmask]>wrote:

> Hello,
>
> I am looking for a test suite of functions that I can use to evaluate
> an evolutionary algorithm I have implemented.
>
> I am interested in functions that have their input ranges (search space)
> well defined along with their maxima/minima.
>
> I keep coming across functions in various articles and books, but many
> often lack this information assuming it's obvious, or just well known. Some
> will only state the maximum or the minimum, others leave the search space
> unspecified.
>
> If there is list of functions with this information available, it would be
> perfect.
>
> Additional recommendations as to what would make a good set of 5 - 10
> functions to include in a test suite would be greatly appreciated.
>
> Thank you,
>
> Esmail
>