A new modification to the genetic algorithm is presented which is specifically designed to increase the rate of evolution on fitness functions with high degrees of neutrality (mutations that do not change the individual's fitness). Instead of allowing random genetic drift to occur when most of the population has reached the same fitness, the "reproduction fitness" of individuals is set to their distance from the population centroid. This has the theoretical effect of spreading the population quickly across the neutral network, and thus finding regions of higher fitness more quickly than it mould otherwise. A series of experiments is described which shows a significant improvement in using this method on the NKp family of fitness functions, and show that this improvement is correlated with the degree of neutrality.
[1]
Christian M. Reidys,et al.
Evolutionary Dynamics and Optimization: Neutral Networks as Model-Landscapes for RNA Secondary-Structure Folding-Landscapes
,
1995,
ECAL.
[2]
M. Huynen,et al.
Smoothness within ruggedness: the role of neutrality in adaptation.
,
1996,
Proceedings of the National Academy of Sciences of the United States of America.
[3]
Inman Harvey,et al.
Through the Labyrinth Evolution Finds a Way: A Silicon Ridge
,
1996,
ICES.
[4]
L. Barnett.
TANGLED WEBS Evolutionary Dynamics on Fitness Landscapes with Neutrality
,
1997
.
[5]
Christian M. Reidys,et al.
Neutrality in fitness landscapes
,
2001,
Appl. Math. Comput..