New selection method to improve the population diversity in genetic algorithms

We present a new method of selection, in order to improve the population diversity in the genotype distribution, in genetic algorithms (GAs). The problem of maintaining of the population diversity is very important in designing genetic operators, when GAs are applied to optimization problems. Therefore, we propose two types of new selection operators based on the correlations between individuals´ genotypes, for improving the population diversity. The first operator is a new type of selection for reproduction, namely the correlative tournament selection. The second operator is a new type of selection for survival, namely correlative family-based selection. We have applied our GA to two different problems: Royal road problems, and Knapsack problems with non-stationary environments. We have compared our method with the other representative GA model, and have shown the effectiveness of the proposed GA models