@InProceedings{FurtadoCampMaca:2012:DaAsNe,
author = "Furtado, Helaine C. M. and Campos Velho, Haroldo F. de and Macau,
Elbert E. N.",
affiliation = "{} and {Instituto Nacional de Pesquisas Espaciais (INPE)} and
{Instituto Nacional de Pesquisas Espaciais (INPE)}",
title = "Data assimilation by neural network emulating representer method
applied to the wave equation",
booktitle = "Proceedings...",
year = "2012",
pages = "476--484",
organization = "International Symposium on Uncertainty Quantification and
Stochastic Modeling, 1.",
keywords = "data assimilation, neural network, variational method, representer
method, wave equation.",
abstract = "Description of a physical phenomenon through differential
equations has errors involved, since the mathematical model is
always an approximation of reality. For an operational prediction
system, one strategy to deal with uncertainties from the modeling
and observation errors is to add some information from the real
dynamics into mathematical model. This aditional information
consists of observations on the phenomenon. However, the
observational data insertion should be done carefully, for
avoiding a worse performance of the prediction. Technical data
assimilation are tools to combine data from physical-mathematics
model with observational data to obtain a better forecast. Two
data assimilation methods are compared here: the Kalman Filter
method, and artificial neural network. Artificial neural networks
appear as a novel method in the context for data assimilation. The
performance of the methods is evaluated under application to wave
propagation model (Bennet,2002).",
conference-location = "S{\~a}o Sebasti{\~a}o, SP",
conference-year = "Feb. 26th to Mar. 2nd, 2012",
issn = "2238-1007",
targetfile = "71-Helaine.pdf",
urlaccessdate = "29 jun. 2024"
}