REVIEW AND APPLICATION OF NON-LINEAR OPTIMISATION METHODS FOR TRAINING OF MULTILAYER PERCEPTRON (PREPRINT IBRAE-1999-09)

Preprint IBRAE-99-09

Timonin V.A., Demyanov V.V., Kanevski M.F. 

Application of different methods of non-linear optimisation is considered for training of multilayer perceptron ANN. Stochastic, genetic and deterministic optimisation algorithms are considered. Different methods are applied to the real data on the radioactive soil contamination of the Chernobyl fallout. 

Bibliographical reference

Timonin V.A., Demyanov V.V., Kanevski M.F. REVIEW AND APPLICATION OF NON-LINEAR OPTIMISATION METHODS FOR TRAINING OF MULTILAYER PERCEPTRON. Preprint IBRAE-99-09. Moscow: Nuclear Safety Institute. November 1999. 28 p. — Refs.: 15 items.