Exploring nonlinear relationships in chemical data using kernel-based methods |
| |
Authors: | Dong-Sheng CaoYi-Zeng Liang Qing-Song Xu Qian-Nan HuLiang-Xiao Zhang Guang-Hui Fu |
| |
Affiliation: | a Research center of modernization of traditional Chinese medicines, Central South University, Changsha 410083, PR Chinab School of Mathematical Sciences and Computing Technology, Central South University, Changsha 410083, PR Chinac Systems Drug Design Laboratory, College of Pharmacy, Wuhan University, Wuhan 430071, PR China |
| |
Abstract: | Kernel methods, in particular support vector machines, have been further extended into a new class of methods, which could effectively solve nonlinear problems in chemistry by using simple linear transformation. In fact, the kernel function used in kernel methods might be regarded as a general protocol to deal with nonlinear data in chemistry. In this paper, the basic idea and modularity of kernel methods, together with some simple examples, are discussed in detail to give an in-depth understanding for kernel methods. Three key ingredients of kernel methods, namely dual form, nonlinear mapping and kernel function, provide a consistent framework of kernel-based algorithms. The modularity of kernel methods allows linear algorithms to combine with any kernel function. Thus, some commonly used chemometric algorithms are easily extended to their kernel versions. |
| |
Keywords: | Kernel methods Dual solution Support vector machines (SVMs) Kernel principal component analysis (KPCA) Kernel partial least squares (KPLS) Kernel Fisher discriminant analysis (KFDA) |
本文献已被 ScienceDirect 等数据库收录! |
|