An application of neural net technology to surveillance information correlation and battle outcome prediction

The author describes a three-layer probabilistic feed-forward neural network that uses sums of Gaussian distributions to estimate the probability density function for a training data set. She shows how this trained network can be used to classify new data sets and to provide a probability associated with each classification. The method has been applied successfully to two separate electronic intelligence emitter correlation problems (hull-to-emitter and land-based emitter correlation). Each of these applications achieved a high degree of accuracy in identifying the correct emitter among many possible emitters about 200000 times faster than the standard back-propagation neural network technique. To show the versatility of the probabilistic neural network for performing optimally any classification problem, an application to battle outcome prediction is described