Título: Redes Compostas por Blocos de Regressões Sigmóides Não-lineares: Uma eficiente rede de alta ordem com ap\licações na previsão de séries temporais
Autores: Valença, Mêuser; Ludermir, Teresa
Resumo: This paper developed a new class of higher order feedforward neural networks, the Non-linear Sigmoidal Regression Blocks Networks (NSRBN), which is able to approximate any continuous function defined over a compact set. These networks are constructive since they deploy an constructive incremental learning algorithm. This algorithm is responsible for the selection of the optimal complexity model, that is, for the network’s architecture definition. Simulation results of forecasting was emphasized, specially in the area of water resources, which was the main motivation of this paper.
Código DOI: 10.21528/CBRN2001-015
Artigo em pdf: 5cbrn_015.pdf
Arquivo BibTex: 5cbrn_015.bib