Papers by Isneri Talavera
Quimiometrix II es un sistema de herramientas concebido para el preprocesamiento, exploración, cl... more Quimiometrix II es un sistema de herramientas concebido para el preprocesamiento, exploración, clasificación y calibración de datos químicos y bioquímicos. Los métodos empleados están sustentados sobre las técnicas más novedosas para el reconocimiento estadístico de patrones. Se implementa en una arquitectura para el desarrollo de software científico basada en Plugin que permite la instalación, actualización y control de errores de manera automatizada. En esta versión se incorporan algoritmos novedosos propios para la calibración multivariante, así como una librería matemática que los implementa (Quimilap). Se presenta un caso de estudio donde se muestra el empleo del software. Palabras clave: análisis exploratorio de datos, clasificación, calibración multivariada.
Signal and image alignment during the application of functional data analysis. Practical examples of chemometrics and biometrics [Alineación de senales e imágenes durante la aplicación del análisis de datos funcionales. Ejemplos prácticos de senales e imágenes quimiometricas y biometricas]
Revista Cubana de …, 2011
Proyecto académico sin fines de lucro, desarrollado bajo la iniciativa de acceso abierto
La técnica analítica de cromatografía de capa fina es una de las más utilizadas actualmente para ... more La técnica analítica de cromatografía de capa fina es una de las más utilizadas actualmente para el análisis de sustancias. Este trabajo se enfoca en el análisis automático de las imágenes de dichas placas mediante métodos de procesamiento de imágenes y reconocimiento de patrones. El método propuesto consta de varias etapas como la captura, detección, recorte y rotación automática de las imágenes de la placa. Luego, se aplican algunos filtros para su mejoramiento y para poder detectar correctamente las manchas de las sustancias de interés. Por último, a partir de esa imagen se realiza la extracción automática de los descriptores de forma y color, y la determinación del valor del Rf para identificar la sustancia mediante el cálculo de la similitud con las sustancias patrones previamente almacenadas. El método presentado es evaluado mediante su aplicación en la identificación de drogas, obteniéndose resultados satisfactorios en la identificación de 44 drogas de abuso.
Quimiometrix II es un sistema de herramientas concebido para el preprocesamiento, exploración, cl... more Quimiometrix II es un sistema de herramientas concebido para el preprocesamiento, exploración, clasificación y calibración de datos químicos y bioquímicos. Los métodos empleados están sustentados sobre las técnicas más novedosas para el reconocimiento estadístico de patrones. Se implementa en una arquitectura para el desarrollo de software científico basada en Plugin que permite la instalación, actualización y control de errores de manera automatizada. En esta versión se incorporan algoritmos novedosos propios para la calibración multivariante, así como una librería matemática que los implementa (Quimilap). Se presenta un caso de estudio donde se muestra el empleo del software. Palabras clave: análisis exploratorio de datos, clasificación, calibración multivariada.
Estuarine, Coastal and Shelf Science, 2008
The statistical technique of functional data analysis (FDA) is applied to a time series analysis ... more The statistical technique of functional data analysis (FDA) is applied to a time series analysis of plankton monitoring data. The analysis is focused on revealing patterns in the seasonal cycle to assess interannual variability of several different taxonomic groups of plankton. Cell concentrations of diatom, dinoflagellate and zooplankton abundances from the Bay of Fundy, Canada provide the observations for analysis. FDA was performed on the log-transformed abundance data as a new approach for treating such types of sparse and noisy data. Differences in the seasonal progression were seen, with peak numbers, timings and abundance levels varying for the three groups as determined by curve registration and higher order derivatives using the objectively fit FDA curves. Nonmetric multidimensional scaling was used to capture seasonal variation among years. These results were further assessed in terms of dominant species and the relationships between groups for different seasons and years. It is anticipated that the easy to use, general and flexible technique of FDA could be applied to a wide variety of marine ecological data that are characterized by missing values and non-Gaussian distributions.
Abstract. The representation of objects by multi-dimensional arrays is widely applied in many res... more Abstract. The representation of objects by multi-dimensional arrays is widely applied in many research areas. Nevertheless, there is a lack of tools to classify data with this structure. In this paper, an approach for classifying objects represented by matrices is intro-duced, based on the advantages and success of the combination strategy, and particularly in the dissimilar-ity representation. A procedure for obtaining the new representation of the data has also been developed, aimed at obtaining a more powerful representation. The proposed approach is evaluated on two three-way data sets. This has been done by comparing the different ways of achieving the new representation, and the traditional vector representation of the ob-jects.
Signal and image alignment during the application of functional data analysis. Practical examples of chemometrics and biometrics [Alineación de senales e imágenes durante la aplicación del análisis de datos funcionales. Ejemplos prácticos de senales e imágenes quimiometricas y biometricas]
Signal and Image Alignment during the Application of Functional Data Analysis. Practical Examples of Chemometrics and Biometrics
A Simulation Study of Functional Density-Based
In this paper a new nonparametric functional regression method is introduced for predicting a sca... more In this paper a new nonparametric functional regression method is introduced for predicting a scalar random variable Y on the basis of a functional random variable X. The prediction has the form of a weighted average of the training data yi, where the weights are determined by the conditional probability density of X given Y = yi, which is assumed to be Gaussian. In this way such a conditional probability density is incorporated as a key information into the estimator. Contrary to some previous approaches, no assumption about the dimensionality of E(XjY = y) or about the distribution of X is required. The new proposal is computationally simple and easy to implement. Its performance is assessed through a simulation study.
In this paper a new nonparametric functional regression method is introduced for predicting a sca... more In this paper a new nonparametric functional regression method is introduced for predicting a scalar random variable Y on the basis of a functional random variable X. The prediction has the form of a weighted average of the training data yi, where the weights are determined by the conditional probability density of X given Y = yi, which is assumed to be Gaussian. In this way such a conditional probability density is incorporated as a key information into the estimator. Contrary to some previous approaches, no assumption about the dimensionality of E(XjY = y) or about the distribution of X is required. The new proposal is computationally simple and easy to implement. Its performance is assessed through a simulation study
Lecture Notes in Computer Science, 2010
In this paper a new nonparametric functional method is introduced for predicting a scalar random ... more In this paper a new nonparametric functional method is introduced for predicting a scalar random variable Y from a functional random variable X. The resulting prediction has the form of a weighted average of the training data set, where the weights are determined by the conditional probability density of X given Y , which is assumed to be Gaussian. In this way such a conditional probability density is incorporated as a key information into the estimator. Contrary to some previous approaches, no assumption about the dimensionality of E(X|Y = y) is required. The new proposal is computationally simple and easy to implement. Its performance is shown through its application to both simulated and real data.
A non-Bayesian predictive approach for statistical calibration
Journal of Statistical Computation and Simulation, 2012
... DOI: 10.1080/00949655.2010.545060 Noslen Hernández a * , Rolando J. Biscay b &amp... more ... DOI: 10.1080/00949655.2010.545060 Noslen Hernández a * , Rolando J. Biscay b & Isneri Talavera a Available online: 09 Jun 2011. ... In some experiments, several observations (m>1) of Y that correspond to the same unknown value x 0 of X are available. ...
Journal of Chemometrics, 2008
The introduction of support vector regression (SVR) and least square support vector machines (LS-... more The introduction of support vector regression (SVR) and least square support vector machines (LS-SVM) methods for regression purposes in the field of chemometrics has provided advantageous alternatives to the existing linear and nonlinear multivariate calibration (MVC) approaches. Relevance vector machines (RVMs) claim the advantages attributed to all the SVM-based methods over many other regression methods. Additionally, it also exhibits advantages over the standard SVM-based ones since: it is not necessary to estimate the error/margin trade-off parameter C and the insensitivity parameter in regression tasks, it is applicable to arbitrary basis functions, the algorithm gives probability estimates seamlessly and offer, additionally, excellent sparseness capabilities, which can result in a simple and robust model for the estimation of different properties. This paper presents the use of RVMs as a nonlinear MVC method capable of dealing with ill-posed problems. To study its behavior, three different chemometric benchmark datasets are considered, including both linear and non-linear solutions. RVM was compared with other calibration approaches reported in the literature. Although RVM performance is comparable with the best results obtained by LS-SVM, the final model achieved is sparser, so the prediction process is faster. Taking into account the other advantages attributed to RVMs, it can be concluded that this technique can be seen as a very promising option to solve nonlinear problems in MVC.
Lecture Notes in Computer Science, 2008
Many regression tasks in practice dispose in low gear instance of digitized functions as predicto... more Many regression tasks in practice dispose in low gear instance of digitized functions as predictor variables. This has motivated the development of regression methods for functional data. In particular, Naradaya-Watson Kernel (NWK) and Radial Basis Function (RBF) estimators have been recently extended to functional nonparametric regression models. However, these methods do not allow for dimensionality reduction. For this purpose, we introduce Support Vector Regression (SVR) methods for functional data. These are formulated in the framework of approximation in reproducing kernel Hilbert spaces. On this general basis, some of its properties are investigated, emphasizing the construction of nonnegative definite kernels on functional spaces. Furthermore, the performance of SVR for functional variables is shown on a real world benchmark spectrometric data set, as well as comparisons with NWK and RBF methods. Good predictions were obtained by these three approaches, but SVR achieved in addition about 20% reduction of dimensionality.
The representation of objects by multi- dimensional arrays is widely applied in many research are... more The representation of objects by multi- dimensional arrays is widely applied in many research areas. Nevertheless, there is a lack of tools to classify data with this structure. In this paper, an approach for classifying objects represented by matrices is intro- duced, based on the advantages and success of the combination strategy, and particularly in the dissimilar- ity representation. A procedure for obtaining the new representation of the data has also been developed, aimed at obtaining a more powerful representation. The proposed approach is evaluated on two three- way data sets. This has been done by comparing the different ways of achieving the new representation, and the traditional vector representation of the ob-
A standard problem in chemometrics is calibration, which aims at predicting a scalar random varia... more A standard problem in chemometrics is calibration, which aims at predicting a scalar random variable Y from a spectrum X. However, if the main problem is to predict Y from X, the physical data generation mechanism is rather that the spectrum X (e.g., an absorbance spectrum) is explained by Y , which is often a chemical variable (e.g., concentration of a substance). Using this physical model X = r(Y ) + ǫ, we propose a nonparametric approach to solve statistical calibration with functional data and to predict Y from X. This approach is based on the conditional probability density of X given Y , f(X|Y ): the proposed predictor takes the form a weighted average of the observed values of Y , where the weights are derived from an nonparametric estimate of f(X|Y ). The estimation of f(X|Y ) is performed with standard nonparametric estimation methods: in the present paper, the proposed estimator is explicitely given in the realistic case where the error ǫ is supposed to fit a Gaussian dist...
2008 19th International Conference on Pattern Recognition, 2008
Conventional multivariate calibration methods have been developed in chemometrics, using linear r... more Conventional multivariate calibration methods have been developed in chemometrics, using linear regression techniques as principal component regression (PCR) and partial least squares (PLS). Nevertheless, nonlinear methods such as neural networks have been also introduced, and more recently support vector (SVR) based methods. This paper presents the application of relevance vector machines regression method (RVMR) as an alternative regression technique based on the Bayesian theory, for the prediction of physical-chemical properties from chemical spectroscopic data of different instrumental sources. In terms of measuring the real effectiveness and generalization capability of this approach, a comparison study of its performance with other known regression techniques are presented. The good results obtained in terms of root mean square error of prediction (RMSEP) in the prediction of properties of interest, combined with the high sparseness capability exhibited, make this approach a good alternative to solve multivariate regression problems in practice.
In this paper a new nonparametric functional regression method is introduced for predicting a sca... more In this paper a new nonparametric functional regression method is introduced for predicting a scalar random variable Y on the basis of a functional random variable X. The prediction has the form of a weighted average of the training data y i , where the weights are determined by the conditional probability density of X given Y = y i , which is assumed to be Gaussian. In this way such a conditional probability density is incorporated as a key information into the estimator. Contrary to some previous approaches, no assumption about the dimensionality of E(X|Y = y) or about the distribution of X is required. The new proposal is computationally simple and easy to implement. Its performance is assessed through a simulation study. RESUMEN En este artículo se introduce un nuevo método de regresión funcional no paramétrico para predecir una variable aleatoria Y de valores reales, sobre la base de una variable aleatoria funcional X. Las predicciones se construyen mediante una promediación ponderada de los datos de entrenamiento y i , donde las ponderaciones están determinadas por la densidad de probabilidad condicional de X dado Y = y i , la cual se supone Gaussiana. De este modo, dicha densidad condicional es incorporada como información clave en el estimador que se propone. Contrariamente a otros enfoques existentes, no se requieren supuestos restrictivos sobre la dimensión de E(X|Y = y) o la distribución de X. La nueva propuesta es computacionalmente simple y fácil de implementar. Su comportamiento es evaluado a través de un estudio de simulación.
Support vector regression for functional data in multivariate calibration problems
Analytica Chimica Acta, 2009
Quantitative analyses involving instrumental signals, such as chromatograms, NIR, and MIR spectra... more Quantitative analyses involving instrumental signals, such as chromatograms, NIR, and MIR spectra have been successfully applied nowadays for the solution of important chemical tasks. Multivariate calibration is very useful for such purposes and the commonly used methods in chemometrics consider each sample spectrum as a sequence of discrete data points. An alternative way to analyze spectral data is to consider each sample as a function, in which a functional data is obtained. Concerning regression, some linear and nonparametric regression methods have been generalized to functional data. This paper proposes the use of the recently introduced method, support vector regression for functional data (FDA-SVR) for the solution of linear and nonlinear multivariate calibration problems. Three different spectral datasets were analyzed and a comparative study was carried out to test its performance with respect to some traditional calibration methods used in chemometrics such as PLS, SVR and LS-SVR. The satisfactory results obtained with FDA-SVR suggest that it can be an effective and promising tool for multivariate calibration tasks.
Uploads
Papers by Isneri Talavera