作者: J.M. Mat�as , A. Vaamonde , J. Taboada , W. Gonz�lez-Manteiga
DOI: 10.1007/S00477-004-0185-5
关键词:
摘要: Critical for an efficient and effective exploitation of a slate mine is to obtain information on its technical quality, in other words, the exploitability potential deposit. We applied support vector machines (SVM) LS-Boosting assessment quality new unexploited area mine, compared results those obtained kriging neural networks. Firstly we analyzed relationship between semi-parametric SVM regularization framework explored different alternatives training these Subsequently, attempt combine both radial projection structures, formulated boosting technique basis function (RBF) networks defined over projections input space (RBFPP). The application techniques our test drilling data demonstrated similar level performance all estimators examined, with main difference occurring shape respective deposit reconstructions. Therefore, choosing techniques, essential aspect will be their ability reproduce morphological characteristics true process. In this paper also evaluate benefits using estimated covariogram as kernel SVMs compare sparsity solutions. show that selection standard ignores variability structure problem produces poorer than when used kernel.