Neural networks for sensor fusion and adaptive classification

Novels types of neural networks are developed and applied to the problem of adaptive classification. These neural networks utilize a maximum likelihood (ML) approach to achieve optimal fusing of all the available information, such as a priori and real-time information coming from a variety of sensors of the same or different types, and utilize fuzzy classification variables to provide for the efficient utilization of incomplete or erroneous data, including numeric and symbolic data. The basic subsystem of these neural networks uses ellipsoids in the classification feature space as classifier building blocks rather than hyperplanes as in conventional neural networks. This novel approach to the architecture and design of neural networks permits the solution of the following problems currently facing neural network technology: (1) optimal utilization of top-down information at all levels of processing; (2) achievement of flexible classifier shapes in the feature space using a parsimonious architecture, and (3) achievement of desired classification performance without excessively large amounts of training data.