Reduce the dimension of the predistortion model coefficients by lasso regression

This paper proposes a method of using lasso regression to estimate the parameters of the predistortion model. Lasso regression can quickly and effectively extract important variables from many variables to simplify the model. In this paper, ten-fold cross-validation is used to confirm the method of...

Full description

Saved in:
Bibliographic Details
Published in2022 IEEE International Conference on Consumer Electronics (ICCE) pp. 1 - 3
Main Authors Yang, Xinrong, Ren, Jijun, Wang, Xing, Song, Qiushuang
Format Conference Proceeding
LanguageEnglish
Published IEEE 07.01.2022
Subjects
Computational complexity
Computational modeling
Conferences
Consumer electronics
Estimation
lasso regression
Predistortion
ten-fold cross-validation
Online AccessGet full text

Cover

More Information
Summary:This paper proposes a method of using lasso regression to estimate the parameters of the predistortion model. Lasso regression can quickly and effectively extract important variables from many variables to simplify the model. In this paper, ten-fold cross-validation is used to confirm the method of lasso regression regularization coefficient. Experiments show that the model coefficients are reduced from the original 125 to 28, a 78% reduction, which can reduce the computational complexity.
ISSN:2158-4001
DOI:10.1109/ICCE53296.2022.9730199