Comparison of selected classifiers in a sea-bottom recognition task

This paper presents a test of three following classifiers: minimum-distance classifier, feed-forward neural network with backpropagation learning scheme and neuro-fuzzy classifier based on NEFCLASS architecture. They have been applied to the sea bottom type classification task over different input spaces. The experiment proved high efficiency of the minimum-distance classifier and the neural network, NEFCLASS performance had been rather poor. Generalization properties of those classifiers are also investigated. Additional conclusions concerning classifiers topology are presented.