邻近点梯度法与交替方向乘子法求解LASSO的性能比较分析
Performance Comparison and Analysis of Proximal Gradient and ADMM for Solving LASSO
陆萍1
作者信息
- 1. 苏州经贸职业技术学院机电与信息学院，苏州 215009
- 折叠
摘要
正则化模型是机器学习、压缩感知与推荐系统等领域的一类重要模型，其具有变量选择与稀疏化处理等功能，可以有效地避免模型的过拟合，完成信号重建或矩阵补全等工作。对稀疏正则化模型进行介绍，分析邻近点梯度算子与交替方向乘子法等最新的求解方法，并对它们的性能进行比较分析。
Abstract
The regularized models play an important role in a lot of fields, such as: machine learning, compressing sensing, recommending system, and so on. With the ability of variable selection and generating sparse solution, the regularized models can avoid over-fitting. They may also be applied to signal reconstruction and matrix completion. Introduces the regularized models, and analyzes two recently developed al-gorithms:proximal gradient and ADMM, compares the performances on solving LASSO.
关键词
正则化模型/LASSO/邻近点算法/交替方向乘子法Key words
Regularized Model/LASSO/Proximal Gradient/ADMM引用本文复制引用
基金项目
江苏省“青蓝工程”骨干教师培养对象，苏州经贸学院院科研课题(KY-ZR1407)
出版年
2015
NETL
NSTL
维普
万方数据