A New Comid-Based Stochastic Coordinate Descent Method for Non-Smooth Losses

TAO Qing; ZHU Ye-lei; LUO Qiang; KONG Kang

doi:10.3969/j.issn.0372-2112.2013.04.024

您当前的位置：

首页 >

文章列表页 >

A New Comid-Based Stochastic Coordinate Descent Method for Non-Smooth Losses

更新时间：2025-07-16

- A New Comid-Based Stochastic Coordinate Descent Method for Non-Smooth Losses
- Acta Electronica Sinica Vol. 41, Issue 4, Pages: 768-775(2013)
- 作者机构：
  
  中国人民解放军陆军军官学院,安徽,合肥,230031
- 作者简介：
- 基金信息：
  
  National Natural Science Foundation of China (No.61273296, No.60975040)
- DOI：10.3969/j.issn.0372-2112.2013.04.024
  CLC： TP301
- Published：2013
- 稿件说明：
移动端阅览
TAO Qing, ZHU Ye-lei, LUO Qiang, et al. A New Comid-Based Stochastic Coordinate Descent Method for Non-Smooth Losses[J]. Acta Electronica Sinica, 2013, 41(4): 768-775.
DOI：

TAO Qing, ZHU Ye-lei, LUO Qiang, et al. A New Comid-Based Stochastic Coordinate Descent Method for Non-Smooth Losses[J]. Acta Electronica Sinica, 2013, 41(4): 768-775. DOI： 10.3969/j.issn.0372-2112.2013.04.024.

摘要

坐标下降方法以简洁的操作流程、低廉的计算代价和快速的实际收敛效果

成为处理大规模优化最有效的方法之一.但目前几乎所有的坐标下降方法都由于子问题解析求解的需要而假设损失函数的光滑性.本文在结构学习的框架下

在采用Comid方法求解随机挑选单变量子问题的基础上

提出了一种新的关于非光滑损失的随机坐标下降方法.理论分析表明本文所提出的算法在一般凸条件下可以得到

(

)的收敛速度

在强凸条件下可以得到

(ln

t/t

)的收敛速度.实验结果表明本文所提出的算法对正则化Hinge损失问题实现了坐标优化预期的效果.

Abstract

Coordinate descent (CD) method is one of the most efficient algorithms in dealing with the large-scale optimization problems for its simple operation

cheap computational cost and practical efficiency.However until now

almost all the-state-of-art CD algorithms require the smoothness assumption of loss functions due to solving the subproblems in closed-form.In this paper

within the structural learning framework

we present a new stochastic CD(SCD)algorithm for non-smooth losses

in which the randomly selected single variable problem is solved using Comid method.Theoretical analysis indicates that the proposed algorithm has an

(

) convergence rate for general convex problems and an

(ln

t/t

) convergence rate for strongly convex problems.The experiments demonstrate the expected efficiency of our SCD algorithms when coping with the L1-regularized Hinge loss problems.

关键词

Keywords

references

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

The Analysis of Convergence Rate of Individual COMID Iterates

A Detection Method for Android Repackaged Malware Based on Sensitive Component Function Call Graph

Research on High-Efficiency Integrated Circuit DFT Technology Based on Machine Learning

DRHA-UIE: An Underwater Image Enhancement Method Based on Dual Residual Hybrid Attention Block

Cognitive Learning: A New Paradigm for Machine Learning in Electromagnetic Spectrum Environment

Related Author

JIANG Ji-yuan

TAO Qing

SHAO Yan-jian

WANG Qun-shan

DU Rui-ying

CHEN Jing

WU Cong

YAN Xi-yu

Related Institution

11th Department, Army Officer Academy of PLA

School of Computer and Information Engineering, Henan Normal University

School of Cyber Science and Engineering, Wuhan University

Key Laboratory of Aerospace Information Security and Trusted Computing, Ministry of Education, Wuhan University

Rizhao Institute of Information Technology, Wuhan University

⁰