摘要
研究数据集被分割并存储于不同处理器时的特征提取和变量选择问题,其中处理器通过某种网络结构相互连接.提出分布式L_(1/2)正则化方法,基于ADMM算法给出分布式L_(1/2)正则化算法,证明了算法的收敛性.算法通过相邻处理器之间完成信息交互,其变量选择结果与数据集不分割时利用L_(1/2)正则化相同.实验表明,所提出的新算法有效、实用,适合于分布式存储数据处理.
This paper focuses on the feature extraction and variable selection of massive data which is divided and stored in different linked computers, and studies the distributed L1/2 regular- ization. Based on Alternating Direction Method of Multipliers algorithm(ADMM), distributed L1/2 regularization algorithm which communicates information between the neighborhood computers has been proposed and the convergence of the algorithm has been proved. The variable selection results of the approach axe the same with the entire data set by using L1/2 regularization. Numerical studies show that this method is both effective and practical which performs well in distributed data analysis.
出处
《高校应用数学学报(A辑)》
CSCD
北大核心
2017年第3期332-342,共11页
Applied Mathematics A Journal of Chinese Universities(Ser.A)
基金
国家自然科学基金(11571011)