Continuous Optimization Schemes for Fuzzy Classification

Two approaches are developed, which are suitable for the optimization of a fuzzy classification scheme through the formation of appropriate space-filling clusters. The first approach is based on the analog Hopfield neural network, while the second one uses real-encoded genetic optimization. Experimental results concerning difficult classification problems show that both proposed approaches are very successful in generating fuzzy partitions and outperform other known algorithms in terms of the correct placement of patterns into partitions.