Modeling of River Discharges using Neural Networks with Structure Determined by Support Vectors  

Oleh:

Fernando T.M.K.G. ; Chan C.W. ; Jayawardena A.W. ; Choy K.Y.

Jenis: Article from Article
Dalam koleksi: Final Program and Book of Abstracts: The 4th Asian Control Conference, September 25-27, 2002 (Sep. 2002), page 618-623.
Topik: SVNN; AMN; Associative Memory Network; Support Networks; Neural Networks
Fulltext: AC021099.PDF (291.4KB)

Isi artikel

Data driven techniques to model and to predict complex and nonlinear systems are increasingly popular, and several class of techniques have been proposed. Neural networks are simple, but are computation intensive, whilst Associative Memory Networks (AMN) are less computation intensive, and yet good parsimonious models of the underlying processes can be obtained. However, to obtain good generalization results, the structure of the AMN has to suitably chosen. An approach to choose these structures is to use the Support Vectors (SV) obtained from the Support Vector Machines. The advantage of this approach is that it is less subjective, as the SVs are obtained by a constrained optimisation using given data set and error bound. For convenience, this class of AMN is referred to as the Support Vector Neural Networks (SVNN). In this paper, the modelling of river discharges with rainfall as input using the SVNN is presented, from which the nonlinear dynamic relationship between rainfall and river discharges is obtained. The ability of SVNN in predicting river discharges from given rainfall is demonstrated. A useful outcome of this study is that the SVNN can provide early warning of severe river discharges when there are heavy rainfalls.

Opini Anda

Klik untuk menuliskan opini Anda tentang koleksi ini!

Kembali