联合低秩表示与图嵌入的无监督特征选择

大数据应用带来高维数据急剧增加，数据降维已成为重要问题.特征选择降维方法已广泛应用于模式识别领域，近年来提出了许多基于流形学习的特征选择方法，然而这类方法往往容易受到各种噪声影响.对此，本文提出一种联合低秩表示和图嵌入的高效无监督特征选择方法（JLRRGE）.通过低秩表示寻找数据在低秩子空间下的表示，降低噪声的影响从而提高算法的鲁棒性，并通过自适应图嵌入方法，使选择特征保持原有的局部关系.实验结果表明，本文提出算法的分类准确率优于其他对比算法.

Joint Low-Rank Representation and Graph Embedding for Unsupervised Feature Selection

Dimensionality reduction becomes a significant problem due to the proliferation of high dimensional data. For dimensionality reduction, feature selection is more analytical than feature extraction. Therefore feature selection plays an important role in pattern recognition. In recent years, many feature selection methods based on manifold learning have been proposed. However, such methods are susceptible to noise data. Therefore, an efficient unsupervised feature selection method is proposed-joint low-rank representation and graph embedding for unsupervised feature selection (JLRRGE). This method not only finds the low rank structure of the data after selecting feature, which makes the algorithm more robust, but also preserves the local manifold structure of the data through the adaptive graph embedding. The experimental results show that the proposed algorithm is superior to other compared methods.