加入收藏
协同高校
浙江大学 清华大学 上海交通大学

A Variance Minimization Criterion to Feature Selection Using Laplacian Regularization

编辑:CYBER日期:2012-11-29 访问次数:815
A Variance Minimization Criterion to Feature Selection Using Laplacian Regularization
作者:He, Xiaofei;  Ji, Ming;  Zhang, Chiyuan;  Bao, Hujun
来源:IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE   卷: 33    期: 10    页: 2013-2025   2011
本文考虑非监督的特征选择问题,基于拉普拉斯正则化最小二乘法提出了两种全新的特征选择算法。新方法的策略是使得参数的协方差矩阵达到最小,从而使得在未知样本上的期望误差最小。分别用迹和行列式两种算子来衡量协方差矩阵的大小,并提出了高效的计算方法。实际数据上的大量实验表明了该算法的优越性。
 
In this paper, we consider the feature selection problem in unsupervised learning scenarios. Based on Laplacian regularized least squares, we propose two novel feature selection algorithms and  select those features such that the size of the parameter covariance matrix of the regularized regression model is minimized. We use trace and determinant operators to measure the size of the covariance matrix. Efficient computational schemes are also introduced to solve the corresponding optimization problems. Extensive experimental results over various real-life data sets have demonstrated the superiority of the proposed algorithms.