文献类型：期刊文献

英文题名：A novel multiple Nystr?m-approximating kernel discriminant analysis

作者：Wang, Zhe[1]; Jie, Wenbo[1]; Gao, Daqi[1]

机构：[1] Department of Computer Science and Engineering, East China University of Science and Technology, Shanghai 200237, China

年份：2013

卷号：119

起止页码：385

外文期刊名：Neurocomputing

收录：EI(收录号：20133416640475)

语种：英文

外文关键词：Eigenvalues and eigenfunctions - Computational efficiency - Discriminant analysis - Pattern recognition

摘要：Multiple Kernel Discriminant Analysis (MKDA) adopts an ensemble of multiple kernel matrices Kis and is supposed to be more flexible and effective than the original Kernel Discriminant Analysis (KDA). However, with n training samples and p kernel matrices Kis, MKDA employs pn2 space units for all the Kis in the optimizing process and simultaneously depends on its solving techniques to handle the optimization problem, which would cause a large space and computational complexity and limit the efficiency and applicability. In order to mitigate this problem, this manuscript adopts the Nystr?m method approximating Ki and therefore develops a novel Multiple Nystr?m-Approximating Kernel Discriminant Analysis (MNKDA). In practice, the proposed MNKDA first adopts m (m{precedes above singleline equals sign}?n) samples to generate an approximating kernel matrix K~i for each Ki and forms an ensemble matrix G=∑i=1pμiK~i. Then, MNKDA directly applies the eigenvalue decomposition onto the Nystr?m-based ensemble matrix G and reformulates the proposed discriminant analysis as an eigenvalue problem. The experimental results show that the proposed method can achieve an effective and efficient performance than the classical MKDA. The advantages of the proposed MNKDA are (1) expressing the formulation as an eigenvalue problem resolution instead of using commercial softwares; (2) decreasing the space complexity from O(pn2) to O(n2) and mitigating the computational complexity from O(n3) to O(pmn2); and (3) providing an alternative multiple kernel learning technique and inheriting the advantage of multiple kernel learning. ? 2013 Elsevier B.V.