Simulated Binary Crossover (SBX) recombinator.

The Simulated Binary Crossover was first proposed by [1]. It i suited for float representation only and creates two offspring. Given parents \(p_1, p_2\) the offspring are generated as \(c_{1/2} = \bar{x} \pm \frac{1}{2}\beta(p_2 - p_1)\) where \(\bar{x} = \frac{1}{2}(p_1 + p_2), p_2 > p_1\). This way \(\bar{c} = \bar{x}\) is assured.

recSBX(inds, eta = 5, p = 1, lower, upper)

Arguments

inds

[numeric] Parents, i.e., list of exactly two numeric vectors of equal length.
eta

[numeric(1)] Parameter eta, i.e., the distance parameters of the crossover distribution.
p

[numeric(1)] Crossover probability for each gene. Default is 1.0.
lower

[numeric] Vector of minimal values for each parameter of the decision space.
upper

[numeric] Vector of maximal values for each parameter of the decision space.

Value

[ecr_recombinator]

Note

This is the default recombination operator used in the NSGA-II EMOA (see nsga2).

References

[1] Deb, K. and Agrawal, R. B. (1995). Simulated binary crossover for continuous search space. Complex Systems 9(2), 115-148.

See also

Other recombinators: recCrossover, recIntermediate, recOX, recPMX, recUnifCrossover