Optimising Neural Networks for Land Use Classification

In this paper we present a fully automatic and computationally efficient algorithm for optimising multilayer perceptron classifiers. The approach involves two procedures: adaptation (training) and selection. The first procedure adaptively changes the weights of the network. The selection procedure performs the elimination of some of the hidden units (weights). By iteratively combining these two procedures we achieve a controlled way of training and modifying neural networks, which balances accuracy, learning time, and complexity of the resulting network. We demonstrate our method on the problem of multispectral Landsat image classification. We compare our results with a hand designed multi-layer perceptron and a Gaussian maximum likelihood classifier on the same data. Our method produces a better classification accuracy with a smaller number of hidden units than the hand designed network.