Kernel principal component analysis
Advances in kernel methods
Kernel Methods for Pattern Analysis
Kernel Methods for Pattern Analysis
An efficient kernel matrix evaluation measure
Pattern Recognition
Input space versus feature space in kernel-based methods
IEEE Transactions on Neural Networks
Optimizing the kernel in the empirical feature space
IEEE Transactions on Neural Networks
| Hi-index | 0.00 |
Choosing appropriate values for kernel parameters is one of the key problems in many kernel-based methods. Beside common used cross validation method which is time-consuming, another kind of rapid methods using kernel matrix evaluation criteria such as Kernel Target Alignment(KTA) and Feature space-based kernel matrix evaluation measurement(FSM) criteria were proposed by researchers. However, we find KTA and FSM maybe failing in learning Gaussian kernel parameter in the case of small sampling size and tend to obtain an overyt solution. In this paper, a novel approach is proposed to learn Gaussian the kernel parameter which works in reproducing kernel mapping space and can avoid above problem. Experiments on real-world datasets show that the proposed approach using the two proposed criteria in this paper works well on learning Gaussian kernel parameter.