Discriminative Feature Fusion Based on Extensions of PCA
TAN Jing-Dong1,2,3, SU Ya-Ru2,3, WANG Ru-Jing2,3
1.School of Mathematics,Hefei University of Technology,Hefei 230009 2.Laboratory of Intelligent Decision,Institute of Intelligent Machines,Chinese Academy of Sciences,Hefei 230031 3.Department of Automation,University of Science and Technology of China,Hefei 230027
Abstract Two methods for dimensionality reduction, principal component discriminative analysis and kernel principal component discriminative analysis, are proposed. Based on the theory of principal component analysis and maximum margin criterion, a multi-objective project model is constructed to formalize the goals for feature fusion. Then, it is transformed into a single-objective cost function for the projection, and the optimal linear mapping is obtained through optimizing this cost function. Additionally, the nearly diagonal block kernel matrix is divided into c kernel matrixes (c is the number of classes in dataset), and eigen-decomposition method is used to solve their d principal vectors. Through the process of vector algebra, a combined mapping α is obtained. When the original kernel matrix K is projected on α, the inner-class information is optimally preserved. The experimental results show their validity.
[1] Hughes N P,Tarassenko L.Novel Signal Shape Descriptors through Wavelet Transforms and Dimensionality Reduction.Proc of SPIE,2003,5207: 763-773 [2] Wang Jue,Zhou Zhihua,Zhou Aoying.Machine Learning and Its Applications.Beijing,China: Tsinghua University Press,2006 (in Chinese) (王 珏,周志华,周傲英.机器学习及其应用,北京:清华大学出版社,2006) [3] Li Haifeng,Jiang Tao,Zhang Keshu.Efficient and Robust Feature Fusion by Maximum Margin Criterion.IEEE Trans on Neural Networks,2006,17(1): 157-165 [4] Li Bo,Zheng Chunhou,Huang Deshuang.Locally Linear Discriminant Embedding: An Efficient Method for Face Recognition.Pattern Recognition,2008,41(12): 3813-3821 [5] Roweis S T.Nonlinear Dimensionality Reduction by Locally Linear Embedding.Science,2000,290(5500): 2323-2326 [6] Shawe-Taylor J,Cristianini N.Kernel Methods for Pattern Analysis.Cambridge,UK: Cambridge University Press,2004 [7] Chatfield C,Collins A J.Introduction to Multivariate Analysis.London,UK: Chapman and Hall,1980 [8] Tan Jingdong.A Study of Some Crucial Algorithms for Text Mining.Ph.D Dissertation.Hefei,China: University of Science and Technology of China,2010 (in Chinese) (檀敬东.文本挖掘的若干关键算法研究.博士学位论文.合肥:中国科学技术大学,2010) [9] Zhang Huaping,Yu Hongkui,Xiong Deyi,et al.HHMM-Based Chinese Lexical Analyzer ICTCLAS // Proc of the 2nd SIGHAN Workshop on Chinese Language Processing.Sapporo,Japan,2003: 184-187 [10] Salton G,McGill M J.Introduction to Modern Information Retrieval.New York,USA: McGraw-Hill,1986 [11] Wang Qinggang,LI Jianwei.An Improved PCA Algorithm with Local Structure Preserving.Pattern Recognition and Artificial Intelligence,2009,22(3): 388-392 (in Chinese) (王庆刚,李见为.具有局部结构保留性质的PCA改进算法.模式识别与人工智能,2009,22(3): 388-392) [12] Hua Shungang,Zhou Yu,Liu Ting.Thermal Infrared Face Image Recognition Based on PCA and LDA.Pattern Recognition and Artificial Intelligence,2008,21(2): 160-164 (in Chinese) (华顺刚,周 羽,刘 婷.基于PCA与LDA的热红外成像人脸识别.模式识别与人工智能,2008,21(2): 160-164) [13] Barshan E,Ghodsi A,Azimifar Z,et al.Supervised Principal Component Analysis: Visualization,Classification and Regression on Subspaces and Submanifolds.Pattern Recognition,2011,44(7): 1357-1371
2012, Vol. 25 
