Introduction to statistical pattern recognition (2nd ed.)
Introduction to statistical pattern recognition (2nd ed.)
Learning and generalization in radial basis function networks
Neural Computation
Evolutionary algorithms in theory and practice: evolution strategies, evolutionary programming, genetic algorithms
Robust classification systems for imprecise environments
AAAI '98/IAAI '98 Proceedings of the fifteenth national/tenth conference on Artificial intelligence/Innovative applications of artificial intelligence
Regularization in the selection of radial basis function centers
Neural Computation
Evolutionary product-unit neural networks classifiers
Neurocomputing
Hybridization of evolutionary algorithms and local search by means of a clustering method
IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
A novel radial basis function neural network for discriminant analysis
IEEE Transactions on Neural Networks
Hi-index | 0.00 |
The machine learning community has traditionally used the correct classification rate or accuracy to measure the performance of a classifier, generally avoiding the presentation of the sensitivities (i.e. the classification level of each class) in the results, especially in problems with more than two classes. Evolutionary Algorithms (EAs) are powerful global optimization techniques but they are very poor in terms of convergence performance. Consequently, they are frequently combined with Local Search (LS) algorithms that can converge in a few iterations. This paper proposes a new method for hybridizing EAs and LS techniques in a classification context, based on a clustering method that applies the k -means algorithm in the sensitivity space, obtaining groups of individuals that perform similarly for the different classes. Then, a representative of each group is selected and it is improved by a LS procedure. This method is applied in specific stages of the evolutionary process and we consider the minimun sensitivity and the accuracy as the evaluation measures. The proposed method is found to obtain classifiers with a better accuracy for each class than the application of the LS over the best individual of the population.