Abstract
One of the well-known and frequently used crossover operators of the real-coded genetic algorithms (GAs) is the blend crossover (BLX). The BLX utilizes a uniform probability distribution function in the offspring production process. The convergence speed of the BLX is not high and sometimes falls to a local optimum for the searching solution problems. In order to improve the searching performance of the real-coded GA, we propose the use of biased probability distribution functions (BPDF) in the crossover process. The crossover with BPDF frequently produces offspring that are close to the best individuals in the current generation and it is highly likely that these offspring will offerr the best solution to the problem. Furthermore we discuss a mutation that has a constant and extended range that is wider than that of the normal BLX. The mutation is able to conserve the population diversity in the generation field due to its wide range and because it maintains a constant offspring production size range. The crossover with BPDF has the role of faster convergence and the mutation has the role of preventing the trap from falling into the local optimum. Simulations show that the crossover between BPDF and the mutation effectively improves the searching performance of two variable optimization problems and a neural network training problem.
