结合流形结构与柔性嵌入的多标签特征选择

将线性回归模型与流形结构相结合,构成了弱线性多标签特征选择的联合框架。首先,用最小二乘损失函数来学习回归系数矩阵;其次,通过标签流形结构来学习数据特征的权重矩阵;再次,用L2,1-范数来约束回归系数矩阵和特征权重矩阵,这样既能引导稀疏性,又有利于特征选择。此外,设计并证明了具有收敛性的迭代更新算法来解决上述提出的问题。最后,所提出的方法在多个经典多标签数据集上进行了验证,实验结果表明了所提算法的有效性。

Multi-label feature selection based on manifold structure and flexible embedding

The linear regression model and manifold structure are combined to form a weak linear multi-label feature selection framework. Firstly, the least square loss function is used to learn the regression coefficient matrix; secondly, the label manifold structure is used to learn the weight matrix of data features; thirdly, L_2,1-norm is used to constrain the regression coefficient matrix and feature weight matrix, which can guide the sparsity and facilitate feature selection. In addition, an iterative updating algorithm with convergence is designed and proved to solve the above problems. Finally, the proposed method is verified on several classical multi-label datasets, and the experimental results show the effectiveness of the proposed algorithm.