Evolutionary Search Based on Aged Structured Population and Selfish Gene Theory

In this paper it is analysed the application of the selfish gene theory using a fusion of concepts proposed in memetic algorithms and selfish gene algorithms. The selfish gene ideas are applied to hierarchical genetic algorithm (HGA) with age structure previously developed by the author. First of all it is analysed the memetic features mainly the learning procedures of the HGA. Secondly, the selfish gene theory (SG) is applied to age structured population aiming to improve the performance of HGA. The evolution of the age structured population is based on the selfish gene theory where the population can be simply seen as a pool of genes and individual genes fight for the same spot in the chromosome of the genotype. The new operator is denoted by Age parameterised Selfish Gene (ApSG) crossover. It is based on a pseudo-crossover scheme with modified Mating Selection Mechanism and Offspring Generation Mechanism influenced by best alleles in the age-structured virtual population. This population evolves to get the best solution using the interaction between the frequencies of the alleles in a same gene group and by changing these frequencies according to the corresponding fitness of individuals. The offspring individuals are inserted into age-structured virtual and current populations according to Lamarckian learning. Finally it is discussed the effects of different learning procedures in the HGA-SG approaches.