Multi-kernel PCA based high-dimensional images feature reduction
| Autor: | Xu Hongzhe, Zhong Weilu, Fu Baiyang, Zheng Weibin, Wen Ge |
|---|---|
| Rok vydání: | 2011 |
| Předmět: |
business.industry
ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Pattern recognition Kernel principal component analysis Heat kernel signature Polynomial kernel Kernel embedding of distributions Variable kernel density estimation Kernel (statistics) Radial basis function kernel Artificial intelligence Tree kernel business Mathematics |
| Zdroj: | 2011 International Conference on Electric Information and Control Engineering. |
| DOI: | 10.1109/iceice.2011.5778352 |
| Popis: | Parameter selection in the intelligent technology model refers to a lot computation of shape image, and there will be much computation of image feature computing. The traditional parameter selection model can not make the shape that will be straight classified in time. Aiming at the bottleneck of the traditional method in high dimensions, based on analysis each kernel function's advantage this paper raises multi-kernel PCA method. This method mixes Multinomial kernel function, Sigmoid kernel function and Gauss radial basis kernel function, make full use of each kernel function's advantage in high dimension shape parameter reduction; Also, Genetic Algorithm is used to determine the key parameters of multi-kernel model. Last, the multi-kernel PCA method is used in shape image Dimension reduction, effectiveness and excellence are tested and verified. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|