W. B. Langdon and R. Poli
thanks for your prompt reply. Yes, I and the research group I work in (Complex
Systems Group, CoSy) have a www home-page of our own:
http://www.irsip.na.cnr.it/~hotg/index.html
where you can find the researchers' biographies, our current activities,
our research fields, our most recent papers in .ps format, so you can download
them, and some links to sites we think are of interest. There is also
one introductory page on Kolmogorov Complexity, and one paper on the
subject to be presented at ICGA97 very soon.
I hope we can start an exchange of ideas which can be of interest for us both.
As regards your question in your email (fitness equal if length are different?),
well, for the problem I have faced the answer is no; in fact in our case
the fitness takes into account that the solution should be as compact as
possible, so the longer a solution is the more it will be penalised. I think you
could give a look at our paper "A Genetic Programming estimate of Kolmogorov
Complexity", which you can find on our www site. In it, you will find the
form of the fitness function we have used in our case (mainly it consists in
three parts:
- degree of similarity between the output of the program on evaluation and the
final string to be compressed;
- length of the program under examination (penalising term);
- difference in length between output string and target string(penalising term).
By making use of this, we have achieved evolutions in which the programs become
shorter and shorter, yet they all can perform the same actions.
Hoping to hear from you very soon
Ivan
Dr. Ivanoe De Falco
IRSIP - CNR
(Institute for Research on Parallel Information Systems -
National Research Council of Italy)
Via P. Castellino, 111
80131 Naples
Italy
ph: ++39 81 5904222
fax: ++39 81 5608330
email: ivan@irsip.na.cnr.it
www page: http://www.irsip.na.cnr.it/~hotg/index.html