Editorial Office:
Management:
R. S. Oyarzabal
Technical Support:
D. H. Diaz
M. A. Gomez
W. Abrahão
G. Oliveira
Publisher by Knobook Pub
doi: 10.6062/jcis.2019.10.02.0161
(Free PDF)J. Anochi, H.F. Campos Velho and R. H. Torres
Neural network is a technique successfully employed in many ap- plications on several research fields. Despite the potential of a neural network model, its performance is dependent on the definition of the parameters, since the definition of architecture (topology) can significantly influence the training process. Here, a technique for au- tomatic configuration for a neural network is described as an optimization problem combining two different optimization schemes: a mono-objective minimization problem using Multi-Particle Collision Algorithm (MPCA), and a multiobjective minimization problem Nondominated Sorting Genetic Algorithm (NSGA-II). The proposed optimization approaches were tested for the mesoscale seasonal climate prediction for precipitation. The meteorological data were processed by Rough Set Theory to extract relevant information to perform the climate prediction by neural network for the Southeast region of Brazil, with a reduced data set.
metaheuristic, optimization problem, neural networks, climate prediction, mono-objective problem, multiobjective problem.