Abstract:
A new double-stranded encoding scheme
based on the DNA molecule is described.
Double-stranded encodings allow the representation
of more schemata than single stranded
encodings containing the same
number of bits. This allows GAs to search
more efficiently. Experiments showed faster
initial search by double stranded encodings
than with single-stranded genomes. Later,
however, inter strand constraints emerge
that reduce the efficiency of search. Techniques
to alleviate the constraints are discussed.
Use of a double-stranded encoding
allows simple implementation of the inversion
genetic operator. Results show that higher
rates of inversion lead to earlier discovery of
high fitness and increased convergence of the
population towards it.
