This paper presents a multiple artificial neural networks (MANN) method with interaction\nnoise for estimating the occurrence probabilities of different classes at any site in space. The MANN\nconsists of several independent artificial neural networks, the number of which is determined by the\nneighbors around the target location. In the proposed algorithm, the conditional or pre-posterior\n(multi-point) probabilities are viewed as output nodes, which can be estimated by weighted\ncombinations of input nodes: two-point transition probabilities. The occurrence probability of\na certain class at a certain location can be easily computed by the product of output probabilities\nusing Bayes� theorem. Spatial interaction or redundancy information can be measured in the form\nof interaction noises. Prediction results show that the method of MANN with interaction noise has\na higher classification accuracy than the traditional Markov chain random fields (MCRF) model and\ncan successfully preserve small-scale features.
Loading....