W. B. Langdon and R. Poli
I wonder if there any evidence (perhaps from the human genome
project?) that multi-cellular organism's genomes' are still increasing
in size, or have they reached some form of equilibrium?
You are right the question of whether to suppress bloat or not is a
very interesting one. There is an argument that some form of
redundancy is in practise necessary for long term evolution. This
suggests that both natural and artificial evolution progress further
over the long term if neutral changes (ie changes which have no
immediate effect) can accumulate eventually allowing an improvement
that would not be possible in a single step. For this to happen there
must be space in the genome to store them. However with simple static
forms of genetic programming, bloat appears to make the population
very conservative and resistant to genetic changes, ie it appears to
make subsequent evolution more difficult.
A frequently reported argument against allowing bloat is symbolic
regression with fixed training data. Here it has been reported that
allowing bloat encourages over fitting, ie evolving to fit the
training data rather than the (supposed) underlying case.
Doubtless the costs associated with bloated chromosomes are small
compared to engineering simulations based upon them but it appears
that unconstrained bloat will eventually consume all the available
machine resources. Ie in practice eventually a limit is reached.
Nevertheless it is a good question whether there is an optimal
genome size and how big it might be. Thank you.
Bill
W. B. Langdon, Phone +44 121-414-4791
School of Computer Science, Fax +44 121-414-4281
University of Birmingham,
Birmingham. B15 2TT United Kingdom
http://www.cs.bham.ac.uk/~wbl/