A crossover operator using independent component analysis for real-coded genetic algorithms
M.T. Takahashi, Hajime Kita · 2002
For real-coded genetic algorithms, there have been proposed many crossover operators. The blend crossover (BLX-/spl alpha/) proposed by L.J. Eshelman and J.D. Schaffer (1993) shows a good searching ability for separable fitness functions. However, because of its component-wise operation, BLX-/spl alpha/ faces difficulties in the optimization of non-separable fitness functions. This paper proposes a novel crossover operator that combines the BLX-/spl alpha/ with independent component analysis (ICA). By applying the ICA to the population, the coordinate system of the search space is transformed so as to increase the separability of the fitness function, and then the BLX-/spl alpha/ is applied. A computer simulation shows the good searching ability of the proposed method for non-separable fitness functions.