Logistic回归正则化的改进ADMM算法  

An improved ADMM algorithm for the regularization of Logistic regression

作　　者：吉锋瑞 文杰[1] JI Feng-rui;WEN Jie(College of Science,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,Jiangsu,China)

机构地区：[1]南京航空航天大学理学院,江苏南京211106

出　　处：《宝鸡文理学院学报（自然科学版）》2020年第2期16-24,共9页Journal of Baoji University of Arts and Sciences(Natural Science Edition)

基　　金：国家自然科学基金青年科学基金项目(11701279)。

摘　　要：目的研究不同正则项(L1正则项、L2正则项、L1-L2正则项和Huber正则项)下的Logistic回归问题。方法利用子问题线性化的方法为标准ADMM(Alternating Direction Method of Multiplier)算法提供一组初始可行解,并将初始可行解代入ADMM算法。结果与结论数值实验表明本文提出的改进算法在一定程度上缩短了标准ADMM算法的计算时间,提高了标准ADMM算法的计算效率。Purposes—To study the Logistic regression under different regularizations(L1 regular term,L2 regular term,L1-L2 regular term and Huber regular term).Methods—Linearization method of subproblems is used to provide a set of initial feasible solutions for the standard ADMM(Alternating Direction Method of Multiplier)algorithm,and the initial feasible solutions are brought into ADMM algorithm.Result and Conclusion—The numerical experiments show that the improved algorithm reduces the computing time of the standard ADMM algorithm,and improves the computational efficiency of the standard ADMM algorithm.

关 键 词：LOGISTIC回归 ADMM 线性化 正则化 

分 类 号：O221.2[理学—运筹学与控制论]