Applications of self organizing topology networks in geosciences and remote sensing

Discusses and demonstrates the use of neural network (NN) techniques for geoscience and remote sensing applications. NN models have some important intrinsic properties which are advantageous in this context. Some of these properties are the distribution free property, the learning capability, and the ease of parallel processing implementations. The authors first present a novel class of learning algorithms which takes full advantage of the NN modeling power; this class is referred to as self-organizing topology networks. Second, they present a novel algorithm which belongs to this class of networks called the self-organizing neural network, and for comparison they provide several experimental results using this and more traditional algorithms. This novel technique proves superior in performance, learning time and modeling power, and requires fewer prior assumptions.<>