Does the affinity matrix influence the performance of the Locality Preserving Projection algorithm
作者： Elias R Silva , George DC Cavalcanti , Tsang Ing Ren , None
DOI: 10.1109/ICSMC.2010.5642399
关键词:
摘要: Classical feature extraction techniques, like PCA and LDA, do not deal properly with multimodal problems. Such techniques create projections that preserve the structure of original data distribution. Locality Preserving Projection (LPP) is a technique which looks for transformation matrix minimizes changes into after transformation. This local captured by affinity matrix. However, there many ways to calculate this The main aim paper evaluate influence different matrices over LPP accuracy. experiments showed correct choice can lead performance gain. Among analyzed matrices, Local Scaling Nearest Neighbor reached best results.
sci-hub.se 参考文章(7)
Agoston E. Eiben, J. E. Smith, Introduction to evolutionary computing ,(2003)
Fan R K Chung, Spectral Graph Theory ,(1996)
Jun-Bao Li, Jeng-Shyang Pan, Shu-Chuan Chu, Kernel class-wise locality preserving projection Information Sciences. ,vol. 178, pp. 1825- 1835 ,(2008) , 10.1016/J.INS.2007.12.001
Masashi Sugiyama, Dimensionality Reduction of Multimodal Labeled Data by Local Fisher Discriminant Analysis Journal of Machine Learning Research. ,vol. 8, pp. 1027- 1061 ,(2007)
Xiaofei He, Locality Preserving Projections neural information processing systems. ,vol. 16, pp. 153- 160 ,(2003)
David G. Stork, Richard O. Duda, Peter E. Hart, Pattern Classification ,(1973)
Kevin Bache, Moshe Lichman, UCI Machine Learning Repository University of California, School of Information and Computer Science. ,(2007)