Evolving feed-forward neural networks through evolutionary mutation parameters

Abstract

In this paper we show a preliminary work on evolutionary mutation parameters in order to understand whether it is possible or not to skip mutation parameters tuning. In particular, rather than considering mutation parameters as global environmental features, we regard them as endogenous features of the individuals by putting them directly in the genotype. In this way we let the optimal values emerge from the evolutionary process itself. As case study, we apply the proposed methodology to the training of feed-forward neural netwoks on nine classification benchmarks and compare it to other five well established techniques. Results show the effectiveness of the proposed appraoch to get very promising results passing over the boring task of off-line optimal parameters tuning. © Springer-Verlag Berlin Heidelberg 2006.