多模态高维数据关联分析的联合协同回归模型  

JOINT COLLABORATIVE REGRESSION MODEL FOR ASSOCIATION ANALYSIS OF HIGH-DIMENSIONAL MULTI-MODAL DATA

作　　者：王凯明[1] 李荣鹏 肖玉柱[1] 宋学力[1] Wang Kaiming;Li Rongpeng;Xiao Yuzhu;Song Xueli(School of Sciences,Chang’an University,Xi’an 710064,Shaanxi,China)

机构地区：[1]长安大学理学院,陕西西安710064

出　　处：《计算机应用与软件》2022年第8期28-33,共6页Computer Applications and Software

基　　金：长安大学中央高校基本科研业务费专项资金资助项目(310812163504)。

摘　　要：针对具有类间差异特性的多模态高维数据的关联分析问题,提出一种联合协同回归模型,其由回归模型和典型相关分析模型组合而成。应用协同回归模型进行多模态数据之间以及多模态数据与表型变量间的关联分析;利用Fused lasso实现类间数据融合,利用l_(1)范数的稀疏作用得到具有共享模式和类特征模式的稀疏典型向量;使用数据融合方法选择重要特征。通过ROC曲线对比表明,该模型与传统模型相比,显著提高了多模态高数据特征选择的准确度。Aiming at the correlation analysis of multi-modal high-dimensional data with inter-class differences,we propose a joint collaborative regression model,which is composed of regression model and typical correlation analysis model.Collaborative regression model was used to analyze the correlation between multi-modal data and between multi-modal data and phenotypic variables.We adopted fused lasso to realize the data fusion between classes,and applied the sparse function of l_(1)-norm to obtain the sparse typical vector with shared pattern and class characteristic pattern.Feature selection was accomplished by data fusion method.ROC curve shows that,compared with the traditional models,the proposed model significantly improves the accuracy of feature selection of multi-modal high-dimensional data.

关 键 词：典型相关分析 协同回归 特征选择 Fused lasso 稀疏 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]