期刊
INFORMATION SCIENCES
卷 468, 期 -, 页码 14-28出版社
ELSEVIER SCIENCE INC
DOI: 10.1016/j.ins.2018.08.035
关键词
Semi-supervised feature selection; Sparse models; Graph Laplacian; l(2,p)-norm; Regression; Scatter matrix
资金
- research affairs of Yazd University Fund
Semi-supervised feature selection, which exploits both labeled and unlabeled data to select the relevant features, has an important role in many real world applications. Most semi-supervised feature selection methods have been exclusively designed for classification problems. In this paper, a semi-supervised framework based on graph Laplacian and mixed convex and non-convex l(2,p)-norm (0 < p <= 1) regularization is proposed for regression problems. In the proposed framework, a semi-supervised graph Laplacain based scatter matrix constructed for regression problems is used to encode the label information of labeled data and the local structure of both labeled and unlabeled data. To solve the mixed convex and non-convex regularized l(2,p)-norm framework, a unified iterative algorithm is proposed. The convergence of the proposed unified algorithm is theoretically and experimentally proved. To evaluate the performance of the proposed framework for regression problems, we perform extensive experiments on different computational drug design regression datasets. The results demonstrate the superiority of the proposed framework in comparison with other feature selection methods in selecting the relevant features. (C) 2018 Elsevier Inc. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据