Classification of imbalanced remote-sensing data by neural networks

The multilayer perceptron neural network has proved to be a very effective tool for the classification of remote-sensing images. Unfortunately, the training of such a classifier by using data with very different a priori class probabilities (imbalanced data) is very slow. This paper describes a learning technique aimed at speeding up the training of a multilayer perceptron when applied to imbalanced data. The results obtained on an optical remote-sensing data set suggest that not only is the proposed technique effective in terms of training speed but it also allows classification results to be more stable with respect to initial weights. © 1997 Elsevier Science B.V.