Abstract

The multi-label feature selection that is regarded as a special case of multi-task learning has received much attention in recent years. In this paper, we propose a novel robust and pragmatic multi-label feature selection method, in which the joint l_2,1-norm minimizations of loss function and regularization are emphasized. Specifically, the loss function based on the l_2,1-norm is robust to outliers, and the l_2,1-norm regularization selects features across all samples with joint sparsity. Besides, the feature information inherent in the data is used to construct the correlation matrix, which explores the correlation between features so as to remove the redundant features. An efficient algorithm based on the augmented Lagrangian multiplier method is proposed to solve the objective function. The extensive experiments compared with several state-of-the-art methods are performed on the multi-label datasets to show the effectiveness of the proposed method.

中文翻译：

---

通过稀疏的多标签学习进行具有特征关联的鲁棒特征选择

摘要

近年来，被认为是多任务学习的特例的多标签特征选择受到了广泛的关注。在本文中，我们提出了一种新颖的鲁棒且实用的多标签特征选择方法，其中强调了损失函数和正则化的联合l _2,1-范数最小化。具体而言，基于l _2,1-范数的损失函数对异常值具有鲁棒性，而l _2,1-norm正则化选择具有稀疏性的所有样本中的特征。此外，利用数据中固有的特征信息来构造相关矩阵，以探索特征之间的相关性，从而消除冗余特征。提出了一种基于增强拉格朗日乘子法的有效算法来求解目标函数。在多标签数据集上进行了与几种最新方法进行的广泛实验，以证明该方法的有效性。