¹College of Science, Jiangxi University of Science and Technology, Ganzhou, P. R. China
²School of Computer Science and Technology, Xidian University, Xi’an, P. R. China
³Center of Interdisciplinary, Lanzhou University, Lanzhou, P. R. China
⁴Institute of Physics, Chinese Academy of Science, Beijing, P. R. China

---

Abstract—We consider the problem of feature extraction for kernel machines. One of the key challenges in this problem is how to detect discriminative features while mapping features into kernel spaces. In this paper, we propose a novel strategy to quantify the importance of features. Firstly, we derive an informative energy model to quantification of feature difference. Secondly, we move the features in the same class closer and push away those belong to different classes according to the model and derivate its objective function. Finally, gradient learning is employed to maximize this function. Experimental results on real data sets have shown the efficient and effective in dealing with projection and classification.

Index Terms—Kernel methods, nonlinear transformation, feature extraction, gradient learning.

[PDF]

Cite: Songhua Liu, Jiansheng Liu, Caiying Ding, Chaoquan Zhang, "Kernel-based Informative Feature Extraction via Gradient Learning," Journal of Computers vol. 7, no. 11, pp. 2813-2820, 2012.