A bounded exploration approach to constructive algorithms for recurrent neural networks - Archive ouverte HAL Access content directly
Journal Articles Neural Networks, IEEE - INNS - ENNS International Joint Conference on Year : 2000

A bounded exploration approach to constructive algorithms for recurrent neural networks

Abstract

When long-term dependencies are present in a time series, the approximation capabilities of recurrent neural networks are difficult to exploit by gradient descent algorithms. It is easier for such algorithms to find good solutions if one includes connections with time delays in the recurrent networks. One can choose the locations and delays for these connections by the heuristic presented here. As shown on two benchmark problems, this heuristic produces very good results while keeping the total number of connections in the recurrent network to a minimum.
Fichier principal
Vignette du fichier
ijcnn96.pdf (6.15 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-01527874 , version 1 (25-05-2017)

Licence

Attribution - NonCommercial - NoDerivatives - CC BY 4.0

Identifiers

  • HAL Id : hal-01527874 , version 1

Cite

Romuald Boné, Michel Crucianu, Gilles Verley, Jean-Pierre Asselin de Beauville. A bounded exploration approach to constructive algorithms for recurrent neural networks. Neural Networks, IEEE - INNS - ENNS International Joint Conference on, 2000. ⟨hal-01527874⟩
45 View
42 Download

Share

Gmail Facebook Twitter LinkedIn More