Hybrid Self Organizing Neurons and Evolutionary Algorithms for Global Optimization

In this work a new algorithm inspired by the self organizing maps combined with evolutionary algorithms is lined up. A neuron in the map is not evolving by itself but it is the result of the application of an evolutionary algorithm during a set of iterations. This idea really helps to increasing the performance of both self organizing maps and evolutionary algorithms while considered individually. The experiments performed in this research envisage test functions having a single criteria but a high number of dimensions. Comparisons with four other well known metaheuristics for optimization (such as differential evolution, particle swarm optimization, simulated annealing) show the performance and efficiency of the proposed approach.