中北大学学报（自然科学版）2018，Vol.39Issue(1)：14-20,7.DOI:10.3969/j.issn.1673-3193.2018.01.003
对称矩阵填充的线性交替最速下降算法研究
The Linear Alternating Steepest Descent Algorithm Research for Symmetric Matrix Completion
摘要
Abstract
The correlation algorithm of symmetric matrix completion was studied.Through using the property of symmetric matrices can be diagonalized,the symmetric matrix was simply factorized.By finding the derivative of each part,the steepest descent direction was obtained.Along the steepest de-scent direction,combined with the inexact linear search method to get the corresponding optimal step size,and further update the iterative matrix.Finally,by analyzing the error,the symmetric matrix was completed accurately.In theory,the convergence of the algorithm was proved.In experiment,the nu-merical experiment was carried out by taking different sampling density to verify the feasibility and va-lidity of the algorithm.关键词
矩阵填充/对称矩阵/交替最小/梯度下降/非精确线性搜索Key words
matrix completion/symmetric matrix/alternating minimization/gradient descent/inexact linear search分类
数理科学引用本文复制引用
韩如意,王川龙..对称矩阵填充的线性交替最速下降算法研究[J].中北大学学报（自然科学版）,2018,39(1):14-20,7.基金项目
国家自然科学基金资助项目(11371275) （11371275）
