Accession Number : ADA254948


Title :   An Application of Exponential Neural Networks to Event-Train Recognition.


Descriptive Note : Final rept. 5 Jan 1990-8 May 1992,


Corporate Author : WRIGHT LAB WRIGHT-PATTERSON AFB OH


Personal Author(s) : Raeth, Peter G


Full Text : https://apps.dtic.mil/dtic/tr/fulltext/u2/a254948.pdf


Report Date : 08 May 1992


Pagination or Media Count : 105


Abstract : The purpose of this project is to investigate neural networks for specific applications in passive electronic warfare (EW) involving restoration of deinterleaved pulse trains to their original broadcast form. The project took a generic event-train approach and focused on event-train recognition. It was determined that back propagation neural networks did not represent a logistically supportable means of training. Gaussian radial basis functions were found to be far superior. This report is composed of three chapters: (1) summary of early experiments, (2) introduction to exponential neural networks, and (3) application to event-train recognition. Each chapter has its own references. We believe that the goal can be reached and that additional experiments, with greater data volumes, are warranted.


Descriptors :   *ELECTRONIC WARFARE , *NEURAL NETS , *PULSE TRAINS , FUNCTIONS , ELECTRONICS , NETWORKS , PROCESSING , PARALLEL PROCESSING , CLASSIFICATION , APPROACH , RECOGNITION , PULSES , PROBABILITY , THREATS , TRAINING , WARFARE , VOLUME


Subject Categories : Cybernetics
      Countermeasures


Distribution Statement : APPROVED FOR PUBLIC RELEASE