An Object-Oriented Classification of Agricultural Data Based on Neural Networks

Abstract

An object-oriented model, based upon six basic abstract data types, is introduced. This model is used to design neural networks. A comparison between the most common neural topologies and graphs described in the associated mathematical theory proves that a neural network may be modelled by a set of classes carrying metrical attributes and connected by topological links. The same model is used to structure the agricultural data required for neural calculations. Applications concerning classification are briefly described.