Accession Number : ADA269253

Title :   Fast-Training Algorithm for Hidden-Layer Forward-Feed Neural Networks.

Descriptive Note : Summary rept. Mar 90-Mar 92,

Corporate Author : ARMY MISSILE COMMAND REDSTONE ARSENAL AL WEAPONS SCIENCE DIRECTORATE

Personal Author(s) : Pethel, Shawn D. ; Bowden, Charles M. ; Sung, Chi C.

Report Date : JUN 1993

Pagination or Media Count : 27

Abstract : A new training algorithm, based on the matrix pseudoinverse method, is shown to train hidden-layer, forward-feed neural networks with high accuracy in a short time for nonlinear time series prediction. The algorithm is applied to chaotic time series generated from the logistics map and the Mackey-Glass delay differential equation and compared to corresponding results generated using a backpropagation training algorithm. We demonstrate orders of magnitude shorter training time and comparable accuracy with the new algorithm and show forecasting and self-generation for these systems.

Descriptors :   *ALGORITHMS, *NEURAL NETS, *TIME SERIES ANALYSIS, *CHAOS, TRAINING, MAPPING(TRANSFORMATIONS), LYAPUNOV FUNCTIONS, EMBEDDING, TRANSFER FUNCTIONS.

Subject Categories : Cybernetics
      Theoretical Mathematics
      Numerical Mathematics

Distribution Statement : APPROVED FOR PUBLIC RELEASE