crossover operator, offspring, fitness
A new general-purpose crossover operator is proposed. The representation of a candidate solution is slightly perturbed, and the ensuing changes in fitness are calculated. Such fitness changes (for parents) are used in constructing the offspring resulting from crossover. Experiments with several sets of problems demonstrate that this approach leads to rapid increases in average and best fitness, and performs much better than traditional general-purpose crossover operators.
Mohan, Chilukuri K., "Selective Crossover: Towards Fitter Offspring" (1997). Electrical Engineering and Computer Science. Paper 98.