Faster Genetic Algorithm for Network Paths∗

Through analyzing the algorithm presented by Mitsuo.Gen, and taking into account the schema theorem and the architecture block hypothesis of a genetic algorithm, we showed some flaws in Mitsuo.Gen algorithm in this paper. A better genetic algorithm for solving shortest path problems is presented further, which is based on the technology of dynamic coding of the priority of vertex and gene weight. The microevolution strategy is also considered fully in the paper. After putting forward the measure of the importance of a vertex in a network structure and its formula, the real coding priority of a vertex is generated in a non-uniform distribution function with the parameters of the measure of the vertex importance. Flexible fitness functions, elitist genes selection, the mountain climbing method for local optimum and other methods are adopted. It indicates that the time efficiency of the algorithm is higher than Mitsuo.Gen algorithm through a lot of numeric examples.