Computer Engineering and Applications ›› 2012, Vol. 48 ›› Issue (3): 189-191.

• 图形、图像、模式识别 • Previous Articles     Next Articles

Adaptively weighted PCA algorithm

YANG Kairui, MENG Fanrong, LIANG Zhizhen   

  1. School of Computer Science and Technology, China University of Mining and Technology, Xuzhou, Jiangsu 221116, China
  • Received:1900-01-01 Revised:1900-01-01 Online:2012-01-21 Published:2012-01-21



  1. 中国矿业大学 计算机科学与技术学院,江苏 徐州 221116

Abstract: Considering the sensitivity of PCA to outliers, a new adaptive weighted PCA is proposed to improve the robustness. Based on PCA, an optimization model by minimizing the weighted reconstruction error is constructed. Information entropy is introduced to adjust the weight of each sample’s reconstruction error. An iterative optimization algorithm is used to solve the model. Experiment results on Yale face database and UCI data sets show the robustness and recognition of the method.

Key words: feature extraction, Principal Component Analysis(PCA), Weighted Principal Component Analysis(WPCA), reconstruction error, robustness

摘要: 针对传统PCA方法对离群点鲁棒性差的问题,提出了一种具有更高鲁棒性且自适应权值的PCA方法。在PCA方法的基础上建立了一个加权的重建误差和最小模型,通过引入信息熵来调节重建误差的权值;通过交替优化算法迭代求解模型。在Yale人脸库和UCI数据集上的实验表明该方法具有很好的鲁棒性和识别率。

关键词: 特征提取, 主成分分析, 加权主成分分析, 重建误差, 鲁棒性