Á. Méndez Civieta, R. E. Lillo Rodríguez, M. C. Aguilera Morillo
Quantile regression models are tools capable of obtaining different levels of quantile estimators for a response variable. The usage of quantile estimators opposed to conditional mean estimators can provide a much richer insight of data, especially in the case of non-symmetric data. While working on high-dimensional problems, sparse solutions have proved to be very useful, both in terms of performance and in terms of interpretability of the models. While working on real problems it is not unusual finding covariates with a natural grouped structure (one can think, for instance, of genetic datasets). This structure opens the door to the search of sparse solutions both at the
group and within group levels. Sparse group LASSO regularization method (SGL) has proved to solve this problem. In this work we investigate the adaptation of the SGL regularization method for quantile regression models, supporting our analysis with synthetic and real datasets.
Palabras clave / Keywords: sparse-group-lasso, quantile-regression, high-dimension
Programado
Sesión GT02-2: Análisis Multivariante y Clasificación (AMyC-2). Organizadora: Eva Boj del Val
29 de mayo de 2018 12:20
Sala 5
