计算机科学与探索 ›› 2018, Vol. 12 ›› Issue (1): 134-142.DOI: 10.3778/j.issn.1673-9418.1612040

• 人工智能与模式识别 • 上一篇    下一篇

基于特别的特征表示方法的局部线性KNN算法

卞则康1+,王士同1,王宇翔2   

  1. 1. 江南大学 数字媒体学院,江苏 无锡  214122
    2. 中国船舶科学研究中心,江苏 无锡  214082
  • 出版日期:2018-01-01 发布日期:2018-01-09

Locally Linear KNN Method Based on Specific Feature Representation

BIAN Zekang1+, WANG Shitong1, WANG Yuxiang2   

  1. 1. College of Digital Media, Jiangnan University, Wuxi, Jiangsu 214122, China
    2. China Ship Science Research Center, Wuxi, Jiangsu 214082, China
  • Online:2018-01-01 Published:2018-01-09

摘要: 提出了一种特别的特征表示方法,并在此基础上提出了一种基于特别的特征表示方法的局部线性K最近邻算法(locally linear K-nearest neighbor method,L2KNN),并将之应用到人脸识别中。特别的特征表示方法是在传统的稀疏表示的基础上,加入了非负约束,改进了传统的稀疏表示的方法,在目标函数中增加了集群正则化项,然后优化新的目标函数得到一个新的近似的特征表示。L2KNN算法具有最近邻集群效应(clustering effect of nearest neighbors,CENN),不仅可以增强测试样本与同类的训练样本之间的相关性,而且可以增强同类训练样本之间的相关性。L2KNN算法进一步应用到L2KNNc(L2KNN-based classifier)分类器中,并提出一种系数截断的方法增加L2KNNc分类器的泛化性能,进一步提高分类器的分类性能。在人脸数据集上的实验结果证明了上述结论。

关键词: 特别的特征表示, 局部线性K最近邻算法L2KNN, 最近邻集群效应(CENN), 系数截断方法

Abstract: This paper proposes a novel specific feature representation, then develops a locally linear K-nearest neighbor method called L2KNN accordingly with applications to face recognition. The specific feature representation improves upon the traditional sparse representation by adding up the nonnegative constraint, and the corresponding objective function with the group regularization is optimized to derive a novel approximate specific representation. The proposed method L2KNN shows the clustering effect of nearest neighbors (CENN). It not only can enhance the correlation between the test sample and the training samples with the same label, but also can enhance the correlation among the training samples. The L2KNN method is developed into its classifier L2KNNc (L2KNN-based classifier) in which a coefficients’ truncating method is used to improve the generalization capability of L2KNNc. The experimental results on the face data set indicate the above claim about L2KNNc

Key words: specific feature representation, locally linear K-nearest neighbor method L2KNN, clustering effect of nearest neighbors (CENN), coefficients' truncating method