From: Kevin Wampler (wampler@cs.washington.edu)
Date: Tue Nov 30 2004 - 23:24:59 PST
In "The Evolutionary Origin of Complex Features" the authors present a
computer simulation of evolution which suggests that complex traits can
evolve from the mutation of simpler traits. The specific simulation
presented is the out of NAND functions. With this objective function
alone, the evolutionary algorithm had little success, but when simpler,
but related functions are reinforced the EQU function evolved relatively
frequently. It was also found that locally disadvantageous mutations were
sometimes made which helped to achieve a better maximum later in the
evolution. This suggests that complex features could evolve from
mutations on simpler features in biological organisms as well -- a
significant point in the theory of evolution.
Although these results are suggestive of a similar biological mechanism,
the conclusion is still not particularly solid. In comparison to the
complexity of many biological features, the EQU function is very simple.
Although nature, of course, has an immensely greater computational power
relative to a computer, some results suggesting that evolutionary
algorithms scale effectively to more complex solutions with an increase in
computing time would be useful. Providing that a distinct correlation can
be shown between the complexity of evolved features and iterations of the
evolution algorithm, the validity of the approach could be tested by
checking if the complexity of biological features matches what is expected
from a 4 billion year computation with resources like those in nature.
This would be a rather useful result, as it would either indicate that the
computer simulation does not well reflect reality, or show that insofar as
it resembles the crucial features of the simulation, it is reasonable for
biological evolution to produce many of the `irreducibly complex'
features in organisms today.
This archive was generated by hypermail 2.1.6 : Tue Nov 30 2004 - 23:24:59 PST