Smoothing ADMM for Sparse-Penalized Quantile Regression with Non-Convex Penalties

Reza Mirzaeifard, Naveen K.D. Venkategowda, Vinay Chakravarthi Gogineni, Stefan Werner

Research output: Contribution to journalArticleScientificpeer-review

27 Downloads (Pure)

Abstract

This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a <italic>secondary convergence iteration</italic>. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of <inline-formula><tex-math notation="LaTeX">$o({k^{-\frac{1}{4}}})$</tex-math></inline-formula> for the sub-gradient bound of the augmented Lagrangian, where <inline-formula><tex-math notation="LaTeX">$k$</tex-math></inline-formula> denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.

Original languageEnglish
Pages (from-to)213-228
Number of pages16
JournalIEEE Open journal of Signal Processing
Volume5
DOIs
Publication statusPublished - 29 Dec 2023
MoE publication typeA1 Journal article-refereed

Keywords

  • ADMM
  • Convergence
  • Convex functions
  • non-smooth and non-convex penalties
  • Optimization
  • Prediction algorithms
  • Quantile regression
  • Signal processing
  • Signal processing algorithms
  • Smoothing methods
  • sparse learning

Fingerprint

Dive into the research topics of 'Smoothing ADMM for Sparse-Penalized Quantile Regression with Non-Convex Penalties'. Together they form a unique fingerprint.

Cite this