结合网上的资料,细看了两种求解PCA的方式。当进行协方差矩阵上求解特征值时,若矩阵的维数较小,则可以使用传统的求解方式,直接求出协方差矩阵的所有特征值和对应的特征向量。但是如果是用在图片方面,加入一张100*100的图片,特征维度高达10000维,协方差矩阵则是10000*10000这种级别,这将耗费的计算量呈平方级增长。面对这样一个难点,用传统方法求解不现实,因此引出奇异值分解(SVD),利用SVD不仅可以解出PCA的解,而且无需大的计算量。奇异值分解的基本公式 ,其中,
除了主对角线上的元素以外全为0,主对角线上的每个元素都称为奇异值,且已按大小排序。
的列向量即是
的特征向量;
的列向量是
的特征向量。在看这些数学推导的时候,感觉到自己数学基础还是很薄弱,花了不少时间去补数学知识。
PCA降维--参考文献
[1] Jon Shlens. A Tutorial on Principal Component Analysis-Derivation, Discussion and Singular Value Decomposition. 2003
[2] J. Yang, D. Zhang, A. Frangi, and J. Yang. Two-Dimensional PCA: A New Approach to Appearance-Based Face Representation and Recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 26, no. 1, pp. 131–137, 2004.
[3] L. Sirovich and M. Kirby. Low-Dimensional Procedure for Characterization of Human Faces. J. Optical Soc. Am., vol. 4, pp. 519-524, 1987.
[4] M. Kirby and L. Sirovich. Application of the KL Procedure for the Characterization of Human Faces. IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 12, no. 1, pp. 103- 108, Jan. 1990.
[5] Zhang D, Zhou Z H. (2D)2: Two-directional two-dimensional PCA for efficient face representation and recognition[J]. Neurocomputing, 69(1):224-231,2005.
[6] G. E. Hinton, R. R. Salakhutdinov. Reducing the dimensionality of data with neural networks. Science 313 (5786): 504-507, 2006.
[7] Nawaf Hazim Barnouti. Face recognition using pca-bpnn with dct implemented on face94 and grimace databases. International Journal of Computer Applications 142(6):8-13 · May 2016
[8] Sun Y, Wang X, Tang X. Deep Learning Face Representation from Predicting 10,000 Classes[C]// IEEE Conference on Computer Vision and Pattern Recognition. IEEE Computer Society, 2014:1891-1898.