A modular neural architecture for bathythermograms classification

Presents a neural approach for bathythermogram classification. A modular architecture of multi-layer perceptrons (MLP) stemming from a preclassification into five main types of temperature profile is used. The types and classes are issued from a pre-established typology. The performance of this approach is evaluated on a Red Sea profiles database. The results show that the method is efficient but suffers from classes overlapping.<
>