ADANNET: Automatic Design of Artificial Neural Networks by Evolutionary Techniques

Evolutionary Local Search of Fuzzy Rules through a Novel Neuro-Fuzzy Encoding Method

“…Therefore their similarity is both in the genotype and in the phenotype. As we will see later, this property will be useful as we will use this encoding with the Hamming crossover operator (Barrios et al, 2001), which is based on the hypothesis that the Hamming distance between two bit strings is a measure of the difference between the two phenotypes that they represent.…”

“…In GLF neuro-fuzzy architectures are built automatically by means of a genetic algorithm that uses a proposed coding method, which, together with the Hamming crossover operator, obtains high performance (Barrios et al, 2001). Figure 5: The discovery subsystem of GLF…”

“…to concentrate the search (more exploration). For a more detailed description of how the Hamming operator works, we refer readers to (Barrios et al, 2001). Evaluation subsystem.…”

“…The Hamming crossover operator (Barrios et al, 2001) takes into consideration the Hamming distance between a set of progenitor strings in order to generate two offsprings. The main feature of this operator is its ability to generate offspring that can be more or less similar to their parents (in terms of Hamming distance), according to a measure of the genetic diversity present in the population, hence dynamically acting on the exploitation versus exploration capabilities ratio of the algorithm.…”