Direct search as unsupervised training algorithm for neural networks

This paper proposes a novel unsupervised training method, based on direct search optimization technique, which could be successfully employed in the finding the optimal free parameters, e.g. weights and biases, of an artificial neural network (ANN). Benchmark data sets of artificial and real-world problems have been used in experiments that enable a comparison with other optimization methods e.g. genetic algorithm and state-of-the-art classifiers. The results provide evidence of the effectiveness of our method regarding the possibility of finding the optimal values of weights and biases of a multilayer perceptron neural network and constructing an ANN autonomously.