快速随机多核学习分类算法

doi:10.3969/j.issn.1001-2400.2016.01.007

Abstract

Abstract:

Multiple kernel learning (MKL) combines multiple kernels in a convex optimization framework and seeks the best line combination of them. Generally, MKL can get better results than single kernel learning, but heavy computational burden makes MKL impractical. Inspired by the extreme learning machine (ELM), a novel fast MKL method based on the random kernel is proposed. When the framework of ELM is satisfied, the kernel parameters can be given randomly, which produces the random kernel. Thus, the sub-kernel scale is reduced largely, which accelerates the training time and saves the memory. Furthermore, the reduced kernel scale can reduce the error bound of MKL by analyzing the empirical Rademacher complexity of MKL. It gives a theoretical guarantee that the proposed method gets a higher classification accuracy than traditional MKL methods. Experiments indicate that the proposed method uses a faster speed, more small memory and gets better results than several classical fast MKL methods.

Key words: multiple kernel learning, extreme learning machine, random kernel, empirical rademacher complexity

CLC Number:

TP775

SUN Tao;FENG Jie. Fast randommultiple kernel learning for classification[J].J4, 2016, 43(1): 36-40.

References

［1］ KWAK N. Nonlinear Projection Trick in Kernel Methods: an Alternative to the Kernel Trick ［J］. IEEE Transactions on Neural Networks and Learning Systems, 2013, 24(12): 2113-2119.
［2］ SHAO Z F, ZHANG L, ZHOU X R, et al. A Novel Hierarchical Semisupervised SVM for Classification of Hyperspectral Images ［J］. IEEE Geoscience and Remote Sensing Letters, 2014, 11(9): 92-97.
［3］ WON K, SAUNDERS C, PRGEL-BENNETT A. Evolving Fisher Kernels for Biological Sequence Classification ［J］. Evolutionary Computation, 2013, 21(1): 81-105.
［4］李映, 焦李成. 基于核Fisher判别分析的目标识别［J］. 西安电子科技大学学报, 2003, 30(2): 179-182.
LI Ying, JIAO Licheng. Target Recognition Based on Kernel Fisher Discriminant ［J］. Journal of Xidian University, 2003, 30(2): 179-182.
［5］ SUN T, JIAO L C, LIU F, et al. Selective Multiple Kernel Learning for Classification with Ensemble Strategy ［J］. Pattern Recognition, 2013, 46(11): 3081-3090.
［6］ BACH F, LANCKRIET G, JORDAN M. Multiple Kernel Learning, Conic Duality, and the SMO Algorithm ［C］//Proceedings of 21th International Conference on Machine Learning. New York: ACM, 2004: 41-48.
［7］ SONNENBURG S, RATSCH G, SCHAFER C, et al. Large Scale Multiple Kernel Learning ［J］. Journal of Machine Learning Research, 2006, 7(1): 1531-1565.
［8］ BACH F, RAKOTOMAMONJY A, CANU S, et al. Simple MKL ［J］. Journal of Machine Learning Research, 2008, 9: 2491-2521.
［9］ XU Z L, JIN R, YANG H Q, et al. Simple and Efficient Multiple Kernel Learning by Group Lasso ［C］//Proceedings of the International Conference on Machine Learning. London: Elsevier Limited, 2010: 1175-1182.
［10］ KLOFT M, BREFELD U, SONNENBURG S, et al. Lp-Norm Multiple Kernel Learning ［J］. Journal of Machine Learning Research, 2011, 12: 953-997.
［11］ HUANG G B, WANG D H, LAN Y. Extreme Learning Machines: a Survey ［J］. International Journal of Machine Learning and Cybernetics, 2011, 2(2): 107-122.
［12］ CHEN J Y, ZHENG G Z, CHEN H J. ELM-MapReduce: MapReduce Accelerated Extreme Learning Machine for Big Spatial Data Analysis ［C］//IEEE International Conference on Control and Automation. Washington: IEEE Computer Society, 2013: 400-405.
［13］ CORTES C, MOHRI M, ROSTAMIZADEH A. Generalization Bounds for Learning Kernels ［C］//Proceedings of the 27th International Conference on Machine Learning. London: Elsevier Limited, 2010: 247-254.

[1]	WANG Xinyuan,XIAO Song,LI Lei,JIAO Lingling. Robust target tracking algorithm based on the ELM and discriminative correlation filter [J]. Journal of Xidian University, 2019, 46(1): 57-63.
[2]	LIU Yuqing;XIAO Song;LI Lei. Online discriminant based superpixel tracking method [J]. Journal of Xidian University, 2018, 45(3): 13-17+168.

Fast randommultiple kernel learning for classification

PDF (PC)

Like

Knowledge

Abstract

Cite this article

share this article

References

Related Articles 2

Metrics

Comments

Recommended 10