Crossover Operation in Differential Evolution
Crossover Operation in Differential Evolution
where βi,j,n is a real random number uniform in the range [0, 1] and Cr , a
constant in [0, 1], is the crossover probability or crossover rate.
In this paper, the child ci,n is required to be different from any of its parents.
This convention is often used in evolutionary algorithms. If the child duplicates
vi,n, a randomly chosen parameter of xi,n, xi,j,n, will replace the corresponding
parameter of the child ci,n, ci,j,n.
On the other hand, if ci,n inherits no parameter from vi,n and hence no evolution
happens, a randomly chosen parameter of the child ci,n, ci,j,n,
will be replaced by the corresponding parameter of the mutant vi,n, vi,j,n.
In this scheme, an integer r is first randomly chosen from [1,N]. It
is the starting point for exponential crossover. ci,r,n of the child ci,n
is taken from vi,r,n of the mutant vi,n.
Parameters of the child after (in cyclic sense) r depends on a
series of Bernoulli experiments of probability Cr . The mutant will
keep donating its parameters to the child until the Bernoulli
experiment is unsuccessful for the first time or the crossover
length is already N −1. The remaining parameters of the child
come from xi,n.
The probability distribution of crossover
length as their names indicate.
How the child inherits parameters from the
mutant.
a. The child inherits a fraction of the mutant
consecutively (in cyclic sense) through
exponential crossover
b. while inheritance through binomial
crossover is non-consecutive.
The Fortran-style pseudo-code of consecutive
binomial crossover. N Bernoulli experiments of
probability Cr are carried out. The number of
successful Bernoulli experiment is the crossover
length L. A staring point r is then randomly
chosen between 1 and N. ci,n inherits L
parameters of mutant vi,n consecutively (in
cyclic sense) from the starting point r
(including). The remaining parameters of ci,n
come from xi,n.
A series of Bernoulli experiments of probability
Cr are carried out in the same way as described
in consecutive exponential crossover to
determine the crossover length L for non-
consecutive exponential crossover. L
parameters are then randomly chosen from vi,n
and inherited by ci,n. xi,n donates the remaining
parameters of ci,n. The Fortran-style pseudo-
code of non-consecutive exponential crossover.
Submitted by:
Purvi bajoria