Capability of feed-forward neural networks for a chemical evaluation of sediments with diffuse reflectance spectroscopy

Diffuse reflectance spectroscopy (0.4-2.5 mu m) is evaluated as fast and no n-destructive method for the analysis of sediments, characterised by a wide range of mineral constituents. Combined with feed-forward artificial neura l networks (ANNs) this technique is used to estimate quantitatively the che mical composition from the sediments based on a supervised training with on e model. The examined characteristics include contents of inorganic carbon, Fe, S, Al, Si, Ca, K, Mg and calcite. The efficiency of several learning a lgorithms (Backpropagation, Quickprop, Resilient propagation (Rprop), Casca de Correlation (CC)) is investigated. All learning algorithms perform well using principal component (PC) scores of the first derivative spectra as in put for the supervised training. ANNs trained with Quickprop and Rprop prod uced most accurate estimations of the chemical characteristics and the perf ormance was better than for standard multivariate statistical tools (stepwi se multiple linear regression (SMLR), principal component analysis (PCA)). An interpretation of the results is given by a detailed consideration of th e correlation structure among the chemical constituents. (C) 2000 Elsevier Science B.V. All rights reserved.