Sófacles Figueredo Carreiro Soares

Cross-validation for the selection of spectral variables using the successive projections algorithm

This work compares the use of a separate validation set and leave-one-out cross-validation to guide the selection of variables in the Successive Projections Algorithm (SPA) for multivariate calibration. Two case studies involving diesel and corn analysis by NIR spectrometry are presented. A graphical interface for SPA is available at www.ele.ita.br/~kawakami/spa/

A modification of the successive projections algorithm for spectral variable selection in the presence of unknown interferents

This work proposes a modification to the successive projections algorithm (SPA) aimed at selecting spectral variables for multiple linear regression (MLR) in the presence of unknown interferents not included in the calibration data set. The modified algorithm favours the selection of variables in which the effect of the interferent is less pronounced. The proposed procedure can be regarded as an adaptive modelling technique, because the spectral features of the samples to be analyzed are considered in the variable selection process. The advantages of this new approach are demonstrated in two analytical problems, namely (1) ultraviolet-visible spectrometric determination of tartrazine, allure red and sunset yellow in aqueous solutions under the interference of erythrosine, and (2) near-infrared spectrometric determination of ethanol in gasoline under the interference of toluene. In these case studies, the performance of conventional MLR-SPA models is substantially degraded by the presence of the interferent. This problem is circumvented by applying the proposed Adaptive MLR-SPA approach, which results in prediction errors smaller than those obtained by three other multivariate calibration techniques, namely stepwise regression, full-spectrum partial-least-squares (PLS) and PLS with variables selected by a genetic algorithm. An inspection of the variable selection results reveals that the Adaptive approach successfully avoids spectral regions in which the interference is more intense.

Multi-core computation in chemometrics: case studies of voltammetric and NIR spectrometric analyses

A aplicacao de tecnicas quimiometricas sofisticadas a grandes conjuntos de dados tem se tornado possivel devido aos continuos aprimoramentos tecnologicos em computadores comerciais. Recentemente, tais aprimoramentos tem sido obtidos principalmente atraves da introducao de processadores com multiplos nucleos. Contudo, o uso eficiente de hardware com multiplos nucleos requer o desenvolvimento de software apropriado para computacao paralela. Este artigo trata da implementacao de paralelismo empregando o Matlab Parallel Computing Toolbox, que requer somente pequenas modificacoes em codigos quimiometricos ja existentes de modo a explorar os beneficios do processamento em multiplos nucleos. Empregando essa ferramenta de software, mostra-se que implementacoes paralelas podem proporcionar expressivos ganhos computacionais. Em particular, considera-se o problema de selecao de variaveis empregando o algoritmo das projecoes sucessivas e o algoritmo genetico, bem como o uso de validacao cruzada em minimos quadrados parciais. Para ilustracao, duas aplicacoes analiticas sao apresentadas: determinacao de proteina em trigo por espectrometria de reflectância no infravermelho proximo e classificacao de oleos vegetais comestiveis por voltametria de onda quadrada. Empregando as implementacoes propostas para computacao paralela, ganhos computacionais de ate 204% foram obtidos. The application of sophisticated chemometrics techniques to large datasets has been made possible by continuing technological improvements in off-the-shelf computers. Recently, such improvements have been mainly achieved by the introduction of multi-core processors. However, the efficient use of multi-core hardware requires the development of software that properly address parallel computing. This paper is concerned with the implementation of parallelism using the Matlab Parallel Computing Toolbox, which requires only simple modifications to existing chemometrics code in order to exploit the benefits of multi-core processing. By using this software tool, it is shown that parallel implementations may provide substantial computational gains. In particular, the present study considers the problem of variable selection employing the successive projections algorithm and the genetic algorithm, as well as the use of cross-validation in partial least squares. For demonstration, two analytical applications are presented: determination of protein in wheat by near-infrared reflectance spectrometry and classification of edible vegetable oils by square-wave voltammetry. By using the proposed parallel computing implementations, computational gains of up to 204% were obtained.

A portable, inexpensive and microcontrolled spectrophotometer based on white LED as light source and CD media as diffraction grid.

A portable, microcontrolled and low-cost spectrophotometer (MLCS) is proposed. The instrument combines the use of a compact disc (CD) media as diffraction grid and white light-emitting diode (LED) as radiation source. Moreover, it employs a phototransistor with spectral sensitivity in visible region as phototransductor, as well as a programmable interrupt controller (PIC) microcontroller as control unit. The proposed instrument was successfully applied to determination of food colorants (tartrazine, sunset yellow, brilliant blue and allura red) in five synthetics samples and Fe(2+) in six samples of restorative oral solutions. For comparison purpose, two commercial spectrophotometers (HP and Micronal) were employed. The application of the t-paired test at the 95% confidence level revealed that there are not significant differences between the concentration values estimated by the three instruments. Furthermore, a good precision in the analyte concentrations was obtained by using MLCS. The overall relative standard deviation (R.S.D.) of each analyte was smaller than 1.0%. Therefore, the proposed instrument offers an economically viable alternative for spectrophotometric chemical analysis in small routine, research and/or teaching laboratories, because its components are inexpensive and of easy acquisition.

Calibration transfer employing univariate correction and robust regression

This paper proposes a new method for calibration transfer, which was specifically designed to work with isolated variables, rather than the full spectrum or spectral windows. For this purpose, a univariate procedure is initially employed to correct the spectral measurements of the secondary instrument, given a set of transfer samples. A robust regression technique is then used to obtain a model with low sensitivity with respect to the univariate correction residuals. The proposed method is employed in two case studies involving near infrared spectrometric determination of specific mass, research octane number and naphthenes in gasoline, and moisture and oil in corn. In both cases, better calibration transfer results were obtained in comparison with piecewise direct standardization (PDS). The proposed method should be of a particular value for use with application-targeted instruments that monitor only a small set of spectral variables.

A new validation criterion for guiding the selection of variables by the successive projections algorithm in classification problems

A simplification in SPA-LDA is proposed to circumvent the need for separate training and validation sets. The number of degrees of freedom is employed in the cost function to avoid model overfitting. Three examples are presented: classification of coffee, diesel and vegetable oils by using UV-Vis spectrometry, NIR spectrometry and voltammetry, respectively.

Multivariate analysis of the dielectric response of materials modeled using networks of resistors and capacitors

We discuss the modeling of dielectric responses of electromagnetically excited networks which are composed of a mixture of capacitors and resistors. Such networks can be employed as lumped-parameter circuits to model the response of composite materials containing conductive and insulating grains. The dynamics of the excited network systems are studied using a state space model derived from a randomized incidence matrix. Time and frequency domain responses from synthetic data sets generated from state space models are analyzed for the purpose of estimating the fraction of capacitors in the network. Good results were obtained by using either the time-domain response to a pulse excitation or impedance data at selected frequencies. A chemometric framework based on a Successive Projections Algorithm (SPA) enables the construction of multiple linear regression (MLR) models which can efficiently determine the ratio of conductive to insulating components in composite material samples. The proposed method avoids restrictions commonly associated with Archies law, the application of percolation theory or Kohlrausch-Williams-Watts models and is applicable to experimental results generated by either time domain transient spectrometers or continuous-wave instruments. Furthermore, it is quite generic and applicable to tomography, acoustics as well as other spectroscopies such as nuclear magnetic resonance, electron paramagnetic resonance and, therefore, should be of general interest across the dielectrics community.

Two-dimensional linear discriminant analysis for classification of three-way chemical data

The two-dimensional linear discriminant analysis (2D-LDA) algorithm was originally proposed in the context of face image processing for the extraction of features with maximal discriminant power. However, despite its promising performance in image processing tasks, the 2D-LDA algorithm has not yet been used in applications involving chemical data. The present paper bridges this gap by investigating the use of 2D-LDA in classification problems involving three-way spectral data. The investigation was concerned with simulated data, as well as real-life data sets involving the classification of dry-cured Parma ham according to ageing by surface autofluorescence spectrometry and the classification of edible vegetable oils according to feedstock using total synchronous fluorescence spectrometry. The results were compared with those obtained by using the spectral data with no feature extraction, U-PLS-DA (Partial Least Squares Discriminant Analysis applied to the unfolded data), and LDA employing TUCKER-3 or PARAFAC scores. In the simulated data set, all methods yielded a correct classification rate of 100%. However, in the Parma ham and vegetable oil data sets, better classification rates were obtained by using 2D-LDA (86% and 100%), compared with no feature extraction (76% and 77%), U-PLS-DA (81% and 92%), PARAFAC-LDA (76% and 86%) and TUCKER3-LDA (86% and 93%).

Classification of individual cotton seeds with respect to variety using near-infrared hyperspectral imaging

This paper proposes the use of Near Infrared Hyperspectral Imaging (NIR-HSI) as a new strategy for fast and non-destructive classification of cotton seeds with respect to variety. A total of 807 seeds of four different cotton varieties are employed in this study. For classification purposes, each seed is represented by an average spectrum obtained by coaveraging the pixel spectra of the NIR-HSI image. Conventional NIR and VIS-NIR spectra are also employed for comparison. By using Partial-Least-Squares Discriminant Analysis (PLS-DA), correct classification rates of 98.0%, 89.7% and 91.7% were achieved in the NIR-HSI, conventional NIR and conventional VIS-NIR datasets. The superiority of the NIR-HSI system can be ascribed to a more comprehensive scan of the seed area, as compared to the conventional VIS-NIR spectrometer.

An Active Search Method for Finding Objects with Near-Optimal Property Values within a Given Set

This paper proposes an active search method aimed at finding objects with optimal or near-optimal y-property values, on the basis of x-variables obtained by indirect, less costly methods. The proposed method progresses in a sequential manner, starting from a small subset of objects with known y-values. At each iteration, the K-nearest neighbour regression technique is employed to obtain estimates ŷ for the objects with unknown y-values. The object with best ŷ value is then subjected to a direct analysis procedure for evaluation of the y-property. Examples are presented with simulated data, as well as actual quantitative structure-activity relationship (QSAR) and near-infrared (NIR) spectrometry datasets. The QSAR and NIR case studies involve the search for maximal antidepressant activity in a set of arylpiperazine compounds and maximal pulp yield in a set of eucalyptus wood samples, respectively. In all these cases, the active search yielded results closer to the maximal y-value compared to the classical Kennard-Stone algorithm for object selection.