Lasso regularization paths for NARMAX models via coordinate descent
Carregando...
Data
Autor(es)
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal de Minas Gerais
Descrição
Tipo
Artigo de evento
Título alternativo
Primeiro orientador
Membros da banca
Resumo
We propose a new algorithm for estimating NARMAX models with L1 regularization for models represented as a linear combination of basis functions. Due to the L1-norm penalty the Lasso estimation tends to produce some coefficients that are exactly zero and hence gives interpretable models. The novelty of the contribution is the inclusion of error regressors in the Lasso estimation (which yields a nonlinear regression problem). The proposed algorithm uses cyclical coordinate descent to compute the parameters of the NARMAX models for the entire regularization path. It deals with the error terms by updating the regressor matrix along with the parameter vector. In comparative timings we find that the modification does not reduce the computational efficiency of the original algorithm and can provide the most important regressors in very few inexpensive iterations. The method is illustrated for linear and polynomial models by means of two examples.
Abstract
Assunto
Modelos matemáticos, Sistemas lineares
Palavras-chave
Computational modeling , Mathematical model , Optimization , Approximation algorithms , Estimation , Data models , Minimization, NARMAX Model , Linear Model , Error Term , Parameter Vector , Design Matrix , Regression Problem , Polynomial Model , Coordinate Descent , Running Time , Least Squares Regression , Error Model , Minimization Problem , Standard Normal Distribution , Linear Problem , Elastic Net , Vertical Dashed Lines , Optimal Prediction , Least Squares Problem , Coordinate Descent Algorithm , Least Angle Regression , Maximum Lag , Past Values , Monomial , Final Comments , Naive Approach
Citação
Departamento
Curso
Endereço externo
https://ieeexplore.ieee.org/document/8430924