Convergence properties of non-crossover genetic algorithm

The canonical genetic algorithm (CGA) applies selection, crossover and mutation operators to solve difficult optimization problems. This paper introduces a new approach to CGA. It applies only mutation and selection operators. It is a non-crossover genetic algorithm (NCGA). The proof of global convergence of NCGA is presented in this paper. The simulation on the NP-complete traveling salesman problem (TSP) shows that NCGA is much faster than the CGA. In terms of computation efficiency, NCGA is a very promising approach. This paper casts doubt on the need of the crossover operator in GAs.