arrow
Volume 12, Issue 1
Hard Thresholding Regularised Logistic Regression: Theory and Algorithms

Lican Kang, Yanyan Liu, Yuan Luo & Chang Zhu

East Asian J. Appl. Math., 12 (2022), pp. 35-52.

Published online: 2021-10

Export citation
  • Abstract

The hard thresholding regularised logistic regression in high dimensions with larger number of features than samples is considered. The sharp oracle inequality for the global solution is established. If the target signal is detectable, it is proven that with a high probability the estimated and true supports coincide. Starting with the KKT condition, we introduce the primal and dual active sets algorithm for fitting and also consider a sequential version of this algorithm with a warm-start strategy. Simulations and a real data analysis show that SPDAS outperforms LASSO, MCP and SCAD methods in terms of computational efficiency, estimation accuracy, support recovery and classification.

  • AMS Subject Headings

62J12, 62J02, 62J07

  • Copyright

COPYRIGHT: © Global Science Press

  • Email address
  • BibTex
  • RIS
  • TXT
@Article{EAJAM-12-35, author = {Kang , LicanLiu , YanyanLuo , Yuan and Zhu , Chang}, title = {Hard Thresholding Regularised Logistic Regression: Theory and Algorithms}, journal = {East Asian Journal on Applied Mathematics}, year = {2021}, volume = {12}, number = {1}, pages = {35--52}, abstract = {

The hard thresholding regularised logistic regression in high dimensions with larger number of features than samples is considered. The sharp oracle inequality for the global solution is established. If the target signal is detectable, it is proven that with a high probability the estimated and true supports coincide. Starting with the KKT condition, we introduce the primal and dual active sets algorithm for fitting and also consider a sequential version of this algorithm with a warm-start strategy. Simulations and a real data analysis show that SPDAS outperforms LASSO, MCP and SCAD methods in terms of computational efficiency, estimation accuracy, support recovery and classification.

}, issn = {2079-7370}, doi = {https://doi.org/10.4208/eajam.110121.210621}, url = {http://global-sci.org/intro/article_detail/eajam/19919.html} }
TY - JOUR T1 - Hard Thresholding Regularised Logistic Regression: Theory and Algorithms AU - Kang , Lican AU - Liu , Yanyan AU - Luo , Yuan AU - Zhu , Chang JO - East Asian Journal on Applied Mathematics VL - 1 SP - 35 EP - 52 PY - 2021 DA - 2021/10 SN - 12 DO - http://doi.org/10.4208/eajam.110121.210621 UR - https://global-sci.org/intro/article_detail/eajam/19919.html KW - Sparse logistic regression, hard thresholding regularisation, PDAS, SPDAS AB -

The hard thresholding regularised logistic regression in high dimensions with larger number of features than samples is considered. The sharp oracle inequality for the global solution is established. If the target signal is detectable, it is proven that with a high probability the estimated and true supports coincide. Starting with the KKT condition, we introduce the primal and dual active sets algorithm for fitting and also consider a sequential version of this algorithm with a warm-start strategy. Simulations and a real data analysis show that SPDAS outperforms LASSO, MCP and SCAD methods in terms of computational efficiency, estimation accuracy, support recovery and classification.

Lican Kang, Yanyan Liu, Yuan Luo & Chang Zhu. (2021). Hard Thresholding Regularised Logistic Regression: Theory and Algorithms. East Asian Journal on Applied Mathematics. 12 (1). 35-52. doi:10.4208/eajam.110121.210621
Copy to clipboard
The citation has been copied to your clipboard