Optimización de la regresión de mínimos cuadrados parciales con funciones Kernel
Loading...
Download
Official URL
Full text at PDC
Publication date
2021
Defense date
15/01/2021
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad Complutense de Madrid
Citation
Abstract
La regresión de mínimos cuadrados parciales (PLS) es un método lineal que busca predecir un conjunto de variables dependientes a partir de un conjunto de predictores, extrayendo factores ortogonales que maximizan la capacidad predictiva, también llamados componentes. Cuando las estructuras de datos exhiben variaciones no lineales, se recurre a la regresión de mínimos cuadrados parciales con kernel (KPLS), que transforma los conjuntos de datos originales a un espacio de características de dimensionalidad arbitraria donde sea posible la generación de un modelo lineal. Una dificultad recurrente al implementar la regresión KPLS es determinar el número de componentes y los parámetros de la función kernel que maximizan su desempeño..
Partial Least Squares (PLS) regression is a linear method that seeks to predict a set of dependent variables from a set of predictors by extracting orthogonal factors that maximize predictive ability, also called components. When data structures exhibit non-linear variations, Kernel Partial Least Squares (KPLS) regression is used, which transforms the original data sets into an arbitrarily dimensioned feature space where a linear model can be generated. A recurring difficulty in implementing KPLS regression is determining the number of components and the parameters of the kernel function that maximize its performance...
Partial Least Squares (PLS) regression is a linear method that seeks to predict a set of dependent variables from a set of predictors by extracting orthogonal factors that maximize predictive ability, also called components. When data structures exhibit non-linear variations, Kernel Partial Least Squares (KPLS) regression is used, which transforms the original data sets into an arbitrarily dimensioned feature space where a linear model can be generated. A recurring difficulty in implementing KPLS regression is determining the number of components and the parameters of the kernel function that maximize its performance...
Description
Tesis inédita de la Universidad Complutense de Madrid, Facultad de Ciencias Matemáticas, leída el 15-01-2021