论文信息 - Inverse model formulation of partial least-squares regression: a robust neural network approach

Inverse model formulation of partial least-squares regression: a robust neural network approach

The Partial Least-Squares Regression (PLSR) approach to statistical calibration model development has been formulated using an inverse model. The inverse model PLSR algorithm is implemented using the Partial Least Squares neural NETwork (PLSNET) architecture. Generalized neural network learning rules derived from a statistical representation error criterion are presented. These learning rules will accommodate a quadratic optimization criterion, providing the linear solution. Optimization functions which grow less than quadratically can also be used to provide a robust solution when the empirical data contains impulsive and colored noise and outliers. The robust optimization criterion also accounts for the higher-order statistics associated with the input data. The inverse model PLSNET learning rules require fewer mathematical operations per weight update than the forward model robust PLSNET algorithms, resulting in faster convergence in many cases.

Fredric M. Ham | Thomas M. McDowall

[1] I Kostanic,et al. Multivariate determination of glucose concentrations from optimally filtered frequency-warped NIR spectra of human blood serum. , 1996, Physiological measurement.

[2] Juha Karhunen,et al. Generalizations of principal component analysis, optimization problems, and neural networks , 1995, Neural Networks.

[3] Ivica Kostanic,et al. Partial least-squares: Theoretical issues and engineering applications in signal processing , 1996 .

[4] Fredric M. Ham,et al. Robust learning in a partial least-squares neural network , 1997 .

[5] Fredric M. Ham,et al. Robust partial least-squares regression: a modular neural network approach , 1997, Defense, Security, and Sensing.

[6] Andrzej Cichocki,et al. Neural networks for optimization and signal processing , 1993 .

[7] Ivica Kostanic,et al. Partial least-squares regression neural network (PLSNET) with supervised adaptive modular learning , 1996, Defense + Commercial Sensing.

[8] 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 .