Spectrum analysis by recursively pruned extended auto‐associative neural network

A method for automatic analysis of high‐dimensional spectra is presented. The method is based on a so‐termed extended auto‐associative neural network, which is an auto‐associative (bottleneck) neural network with an additional output that corresponds to the response variable of interest. The input of the neural network consists of the absorbance at selected wavelengths. Once the model has been calibrated, the contribution of each input is estimated, the least contributing input and the corresponding output are removed (pruned), and the training procedure is repeated. This procedure leads to a compact non‐linear mapping between the absorbance at a few wavelengths and the response variable of interest. The auto‐associative architecture of the model prevents the overparametrization and overfitting that occur in straightforward mapping between spectra and variable of interest. Overparametrization and overfitting are further reduced by retaining only the inputs required for predicting the variable of interest. The method is illustrated on two case studies involving mid‐infrared absorbance spectra. The results are compared to those obtained with partial least squares (PLS) and principal component analysis (PCA) followed by a neural network. In both case studies, the proposed approach leads to models with fewer parameters and smaller prediction errors. Copyright © 2006 John Wiley & Sons, Ltd.

[1]  A. S. Weigend,et al.  Selecting Input Variables Using Mutual Information and Nonparemetric Density Estimation , 1994 .

[2]  Michael C. Mozer,et al.  Skeletonization: A Technique for Trimming the Fat from a Network via Relevance Assessment , 1988, NIPS.

[3]  Desire L. Massart,et al.  Variable selection for neural networks in multivariate calibration , 1998 .

[4]  Louis W. Glorfeld.  A Methodology for simplification and interpretation of backpropagation-based neural network models , 1996 .

[5]  Igor V. Tetko,et al.  Neural Network Studies, 2. Variable Selection , 1996, J. Chem. Inf. Comput. Sci..

[6]  George W. Irwin,et al.  Nonlinear PLS using Radial Basis Functions , 1997 .

[7]  Desire L. Massart,et al.  The optimal brain surgeon for pruning neural network architecture applied to multivariate calibration , 1998 .

[8]  B. J. Bailey,et al.  Neural Network Models of the Greenhouse Climate , 1994 .

[9]  G. David Garson,et al.  Interpreting neural-network connection weights , 1991 .

[10]  N. Bodor,et al.  Neural network studies: Part 3. Prediction of partition coefficients , 1994 .

[11]  Ravinder Nath,et al.  Determining the saliency of input variables in neural network classifiers , 1997, Comput. Oper. Res..

[12]  M. Kramer.  Nonlinear principal component analysis using autoassociative neural networks , 1991 .

[13]  I. Shmulevich,et al.  Fourier Transform Infrared—Attenuated Total Reflection Nitrate Determination of Soil Pastes Using Principal Component Regression, Partial Least Squares, and Cross-Correlation , 2004, Applied spectroscopy.

[14]  A. J. Morris,et al.  Non-linear projection to latent structures revisited (the neural network PLS algorithm) , 1999 .

[15]  David M. Haaland,et al.  Partial least-squares methods for spectral analyses. 2. Application to simulated and glass spectral data , 1988 .

[16]  S. Wold.  Nonlinear partial least squares modelling II. Spline inner relation , 1992 .

[17]  E. Martin,et al.  Non-linear projection to latent structures revisited: the quadratic PLS algorithm , 1999 .

[18]  P. C. Robert.  Precision agriculture: a challenge for crop nutrition management , 2002 .

[19]  Thomas J. McAvoy,et al.  Nonlinear PLS Modeling Using Neural Networks , 1992 .

[20]  Y. Etzion,et al.  Determination of protein concentration in raw milk by mid-infrared fourier transform infrared/attenuated total reflectance spectroscopy. , 2004, Journal of dairy science.

[21]  Rénato Froidevaux,et al.  On-Line Mid-Infrared Spectroscopic Data and Chemometrics for the Monitoring of an Enzymatic Hydrolysis , 2001 .

[22]  E. V. Thomas,et al.  Partial least-squares methods for spectral analyses. 1. Relation to other quantitative calibration methods and the extraction of qualitative information , 1988 .

[23]  Manfred Morari,et al.  PLS/neural networks , 1992 .

[24]  Edward C. Malthouse,et al.  Nonlinear partial least squares , 1997 .