The significance of border training patterns in classification by a feedforward neural network using back propagation learning

Training patterns vary in their importance in image classification. Consequ ently, the selection and refinement of training sets can have a major impac t on classification accuracy. For classification by a neural network, train ing patterns that lie close to the location of decision boundaries in featu re space may aid the derivation of an accurate classification. The role of such border training patterns and their identification is discussed in rela tion to a series of crop classifications from airborne Thematic Mapper data . It is shown that a neural network trained with a set of border patterns m ag have a lower accuracy of learning but a significantly higher accuracy of generalization than one trained with a set of patterns drawn from the core s of the classes. Unfortunately, conventional training pattern selection an d refinement procedures tend to favour core training patterns. For classifi cation by a neural network, procedures which encourage the inclusion of bor der training patterns should be adopted as this may facilitate the producti on of an accurate classification.