Theoretical analysis of the unimodal normal distribution crossoverfor real-coded genetic algorithms

摘要

For real-coded genetic algorithms, there have been proposed manycrossover operators so far. While they have been evaluated by somebenchmark problems, theoretically clear guidelines or design principlesfor them have not been established yet. This paper, first, discusses theimportance of the distribution and statistics of the offspring yieldedby a crossover operator for its evaluation. Then, from this viewpoint,the unimodal normal distribution crossover (UNDX) developed by Ono etal. (1997) is analyzed. The results of analysis provide us with a clearunderstanding of the characteristics of the UNDX. It is also shown thatthe values of the adjustable parameters of the UNDX tuned empirically isdesirable in the sense that the offspring population inherits thestatistics such as the mean value and the covariance matrix from theparent population