Hybrid neural networks for tactical target recognition

Summary form only given, as follows. Artificial neural network topologies which use both self-organization and supervised learning are discussed. Aberrations in counterpropagation are shown. A hybrid-network is developed and shown to be more efficient than backward error propagation alone for the tactical target recognition problem considered. The hybrid network is demonstrated as an effective tool for solving pattern recognition problems involving ambiguous decision regions. Tactical target data sets are classified using hybrid propagation and the results are compared to conventional backpropagation networks.<<ETX>>