Sparse rational function methods for developing parsimonious multivariate calibrations

Pekka Teppola, Veli-Matti Taavitsainen

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsProfessional

Abstract

The aim is to demonstrate the flexibility and robustness of two new additions to the rational function methods, namely, rational function least angle regression (LARS), rational function least absolute shrinkage and selection operator (LASSO) and elastic net (ENET) methods. Previous work [1, 2] has been based on the methods that use a L2 regularization to cope with collinearity and measurement noise. In this work, we introduce parsimonious methods for rational function modeling that use rather L1 regularization or the combination of L1 and L2 regularization. These methods provide interesting opportunities for model development and robustification. The key point is to specifically illustrate the flexibility of rational function ENET because it bridges the gap between L1 and L2 methods. The former suffers from spectral noise, which is well known in the context of normal LASSO approaches. The latter has been troublesome in numerous occasions as it involves the use of a broad spectral range that is often prone to suffer from unknown spectral interferences in future samples. These points will also be exemplified using two data sets generated by a multipoint NIR instrument with two 5-channel measurement probes and by a chemical imaging NIR spectrometer. We also briefly demonstrate how to interpret and visualize rational function models efficiently. In summary, this work addresses new and interesting directions in developing calibration models in the field of spectroscopy and multipoint measurements.
Original languageEnglish
Title of host publicationProceedings of CAC 2012
Publication statusPublished - 2012
MoE publication typeD3 Professional conference proceedings
EventXXIIth Chemometrics in Analytical Chemistry, CAC 2012 - Budapest, Hungary
Duration: 25 Jun 201229 Jun 2012

Conference

ConferenceXXIIth Chemometrics in Analytical Chemistry, CAC 2012
Abbreviated titleCAC 2012
CountryHungary
CityBudapest
Period25/06/1229/06/12

Fingerprint

Multivariate Calibration
Rational function
Elastic Net
Regularization
Shrinkage
Flexibility
Collinearity
Model Calibration
Operator
Spectrometer
Demonstrate
Spectroscopy
Probe
Regression
Interference
Imaging
Robustness
Angle
Unknown
Modeling

Keywords

  • Generalized ridge regression
  • rational function ridge regression
  • chemometrics
  • PAT
  • chemical imaging

Cite this

Teppola, P., & Taavitsainen, V-M. (2012). Sparse rational function methods for developing parsimonious multivariate calibrations. In Proceedings of CAC 2012
Teppola, Pekka ; Taavitsainen, Veli-Matti. / Sparse rational function methods for developing parsimonious multivariate calibrations. Proceedings of CAC 2012. 2012.
@inproceedings{b22a576813c048528c493f9bde74849f,
title = "Sparse rational function methods for developing parsimonious multivariate calibrations",
abstract = "The aim is to demonstrate the flexibility and robustness of two new additions to the rational function methods, namely, rational function least angle regression (LARS), rational function least absolute shrinkage and selection operator (LASSO) and elastic net (ENET) methods. Previous work [1, 2] has been based on the methods that use a L2 regularization to cope with collinearity and measurement noise. In this work, we introduce parsimonious methods for rational function modeling that use rather L1 regularization or the combination of L1 and L2 regularization. These methods provide interesting opportunities for model development and robustification. The key point is to specifically illustrate the flexibility of rational function ENET because it bridges the gap between L1 and L2 methods. The former suffers from spectral noise, which is well known in the context of normal LASSO approaches. The latter has been troublesome in numerous occasions as it involves the use of a broad spectral range that is often prone to suffer from unknown spectral interferences in future samples. These points will also be exemplified using two data sets generated by a multipoint NIR instrument with two 5-channel measurement probes and by a chemical imaging NIR spectrometer. We also briefly demonstrate how to interpret and visualize rational function models efficiently. In summary, this work addresses new and interesting directions in developing calibration models in the field of spectroscopy and multipoint measurements.",
keywords = "Generalized ridge regression, rational function ridge regression, chemometrics, PAT, chemical imaging",
author = "Pekka Teppola and Veli-Matti Taavitsainen",
year = "2012",
language = "English",
booktitle = "Proceedings of CAC 2012",

}

Teppola, P & Taavitsainen, V-M 2012, Sparse rational function methods for developing parsimonious multivariate calibrations. in Proceedings of CAC 2012. XXIIth Chemometrics in Analytical Chemistry, CAC 2012, Budapest, Hungary, 25/06/12.

Sparse rational function methods for developing parsimonious multivariate calibrations. / Teppola, Pekka; Taavitsainen, Veli-Matti.

Proceedings of CAC 2012. 2012.

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsProfessional

TY - GEN

T1 - Sparse rational function methods for developing parsimonious multivariate calibrations

AU - Teppola, Pekka

AU - Taavitsainen, Veli-Matti

PY - 2012

Y1 - 2012

N2 - The aim is to demonstrate the flexibility and robustness of two new additions to the rational function methods, namely, rational function least angle regression (LARS), rational function least absolute shrinkage and selection operator (LASSO) and elastic net (ENET) methods. Previous work [1, 2] has been based on the methods that use a L2 regularization to cope with collinearity and measurement noise. In this work, we introduce parsimonious methods for rational function modeling that use rather L1 regularization or the combination of L1 and L2 regularization. These methods provide interesting opportunities for model development and robustification. The key point is to specifically illustrate the flexibility of rational function ENET because it bridges the gap between L1 and L2 methods. The former suffers from spectral noise, which is well known in the context of normal LASSO approaches. The latter has been troublesome in numerous occasions as it involves the use of a broad spectral range that is often prone to suffer from unknown spectral interferences in future samples. These points will also be exemplified using two data sets generated by a multipoint NIR instrument with two 5-channel measurement probes and by a chemical imaging NIR spectrometer. We also briefly demonstrate how to interpret and visualize rational function models efficiently. In summary, this work addresses new and interesting directions in developing calibration models in the field of spectroscopy and multipoint measurements.

AB - The aim is to demonstrate the flexibility and robustness of two new additions to the rational function methods, namely, rational function least angle regression (LARS), rational function least absolute shrinkage and selection operator (LASSO) and elastic net (ENET) methods. Previous work [1, 2] has been based on the methods that use a L2 regularization to cope with collinearity and measurement noise. In this work, we introduce parsimonious methods for rational function modeling that use rather L1 regularization or the combination of L1 and L2 regularization. These methods provide interesting opportunities for model development and robustification. The key point is to specifically illustrate the flexibility of rational function ENET because it bridges the gap between L1 and L2 methods. The former suffers from spectral noise, which is well known in the context of normal LASSO approaches. The latter has been troublesome in numerous occasions as it involves the use of a broad spectral range that is often prone to suffer from unknown spectral interferences in future samples. These points will also be exemplified using two data sets generated by a multipoint NIR instrument with two 5-channel measurement probes and by a chemical imaging NIR spectrometer. We also briefly demonstrate how to interpret and visualize rational function models efficiently. In summary, this work addresses new and interesting directions in developing calibration models in the field of spectroscopy and multipoint measurements.

KW - Generalized ridge regression

KW - rational function ridge regression

KW - chemometrics

KW - PAT

KW - chemical imaging

M3 - Conference article in proceedings

BT - Proceedings of CAC 2012

ER -

Teppola P, Taavitsainen V-M. Sparse rational function methods for developing parsimonious multivariate calibrations. In Proceedings of CAC 2012. 2012