Last semester (my first at the
University
of Tennessee) I took an Independent Study (CS 593)
course with
Dr.
Michael Vose. During that course, we tried to optimize
several Genetic algorithm operators for crossover, and the traditional
bit-flip mutation
operator.
As a result for that course, a draft article tentatively called "
An efficient implementation of
crossover and mutation in binary coded GAs" was produced.
In this article, four crossover operators and one mutation are studied,
and speed-ups of up to 4 times are registered. Also, a Chi-square
statistical test is applied to the results to support correctness of
the new implementations against the traditional ones.
The medullar part of this work lies in a versatile random number
generator that allows for arbitrarily distributed pseudo-random numbers
to be generated in only about 3 times the time that is required to
generate an uniformly distributed random number.
The article is currently in final draft, and being checked by Dr. Vose.