Application of artificial neural networks to chemostratigraphy

Abstract

Artificial neural networks, a branch of artificial intelligence, are computer systems formed by a number of simple, highly interconnected processing units that have the ability to learn a set of target vectors from a set of associated input signals. Neural networks learn by self‐adjusting a set of parameters, using some pertinent algorithm to minimize the error between the desired output and network output. We explore the potential of this approach in solving a problem involving classification of geochemical data. The data, taken from the literature, are derived from four late Quaternary zones of volcanic ash of basaltic and rhyolithic origin from the Norwegian Sea. These ash layers span the oxygen isotope zones 1, 5, 7, and 11, respectively (last 420,000 years). The data consist of nine geochemical variables (oxides) determined in each of 183 samples. We employed a three‐layer back propagation neural network to assess its efficiency to optimally differentiate samples from the four ash zones on the basis of their geochemical composition. For comparison, three statistical pattern recognition techniques, linear discriminant analysis, the k‐nearest neighbor (k‐NN) technique, and SIMCA (soft independent modeling of class analogy), were applied to the same data. All of these showed considerably higher error rates than the artificial neural network, indicating that the back propagation network was indeed more powerful in correctly classifying the ash particles to the appropriate zone on the basis of their geochemical composition.

Machine learning classifiers for attributing tephra to source volcanoes: an evaluation of methods for Alaska tephras

“…As such, no performance measure was given. One other early study of statistical cheomstratigraphy, Malmgren and Nordlund (), compared ANN, KNN, and LDA for classifying volcanic ash zones utilizing major oxide geochemistry. Their basic findings agree with ours, indicating high accuracy of ANN on held‐out data (our study = 97.7%, Malmgren and Nordlund () = 90.8%).…”

Accurate classification of juvenile weakfish Cynoscion regalis to estuarine nursery areas based on chemical signatures in otoliths

“…Feed-forward neural networks are becoming increasingly popular in pattern recognition and classification applications in a number of scientific disciplines, includng geology (Malrngren & Nordlund 1996), population genetics (Cornuet et al 1996), and ecology (Culverhouse et al 1992). The neural network that we used consisted of an input matrix (the multivariate elemental data set) that passes to a single hidden layer after a tan-sigmoid transformation with a weight and bias term.…”

Application of artificial neural networks (ANN) to high‐latitude dinocyst assemblages for the reconstruction of past sea‐surface conditions in Arctic and sub‐Arctic seas

“…The term 'b' constitutes a bias, which is an adjustable parameter analogue to constant values in regression equations. The training process, which consists in estimating the adjustable parameters (biases and weights) of an ANN, involves adjustments of the weights in order to find an optimal formal mapping between a set of explicative variables in the input layer and a corresponding set of output variables in the last layer (Malmgren and Nordlund, 1996). A training set with input and output variables, here the relative abundance of dinocyst taxa and the sea-surface parameters, is thus required.…”

“…In the back propagation step, the difference or 'error' between the output predicted from the forward propagation process and the desired (observed) output variables is computed. The error values are used to incrementally adjust the weights between the output layer and the hidden layer according to a learning algorithm (Malmgren and Nordlund, 1996). The forward propagation and back propagation steps are repeated in an iterative procedure until the difference between the output predicted and the desired output is stabilised, which corresponds then to the optimal ANN calibration.…”

“…The ANN approach evolved rapidly in various fields of application, in psychology, medical sciences, image processing and earth sciences, for example (e.g. Anderson and Rosenfeld, 1987;Malmgren and Nordlund, 1996). Applications of the ANN approach to ecology are still rare, although they could provide better results than traditional statistical techniques (Lek et al, 1996;Moatar et al, 1999).…”