Kernel-based maximum correntropy criterion with gradient descent method

Ting Hu

doi:10.3934/cpaa.2020186

Article Contents

2020, Volume 19, Issue 8: 4159-4177. Doi: 10.3934/cpaa.2020186

This issue Previous Article Computing eigenpairs of two-parameter Sturm-Liouville systems using the bivariate sinc-Gauss formula Next Article A numerical method to compute Fisher information for a special case of heterogeneous negative binomial regression

Kernel-based maximum correntropy criterion with gradient descent method

Ting Hu

School of Mathematics and Statistics, Wuhan University, Wuhan, China

Received: August 31, 2019

Revised: November 30, 2019

Published: May 2020

The author is supported by NSFC grant 11671307 and 11571078

Abstract Full Text(HTML) Related Papers Cited by

Abstract

In this paper, we study the convergence of the gradient descent method for the maximum correntropy criterion (MCC) associated with reproducing kernel Hilbert spaces (RKHSs). MCC is widely used in many real-world applications because of its robustness and ability to deal with non-Gaussian impulse noises. In the regression context, we show that the gradient descent iterates of MCC can approximate the target function and derive the capacity-dependent convergence rate by taking a suitable iteration number. Our result can nearly match the optimal convergence rate stated in the previous work, and in which we can see that the scaling parameter is crucial to MCC's approximation ability and robustness property. The novelty of our work lies in a sharp estimate for the norms of the gradient descent iterates and the projection operation on the last iterate.

Keywords:

Mathematics Subject Classification: Primary: 68Q32, 62J02; Secondary: 41A46.

Citation:

Full Text(HTML)

Related Papers

Cited by

References

[1]	N. Aronszajn, Theory of reproducing kernels, Trans. Amer. Math. Soc., 68 (1950), 337-404. doi: 10.2307/1990404.
[2]	R. J. Bessa, V. Miranda and J. Gama, Entropy and correntropy against minimum square error in offline and online three-day ahead wind power forecasting, IEEE Trans. Power Syst., 24 (2009), 1657-1666. doi: 10.1109/TPWRS.2009.2030291.
[3]	D. R. Chen, Q. Wu, Y. Ying and D. X. Zhou, Support vector machine soft margin classifiers: Error analysis, J. Mach. Learn. Res., 5 (2004), 1143-1175.
[4]	M. Debruyne, A. Christmann, M. Hubert and J. A. K. Suykens, Robustness of reweighted least squares kernel based regression, J. Multi. Anal., 101 (2010), 447-463. doi: 10.1016/j.jmva.2009.09.007.
[5]	Y. Feng, J. Fan and J. A. Suykens, A statistical learning approach to modal regression, J. Mach. Learn. Res., 21 (2020), 1-35.
[6]	Y. Feng, X. Huang, S. Lei, Y. Yang and J. A. K. Suykens, Learning with the maximum correntropy criterion induced losses for regression, J. Mach. Learn. Res., 16 (2015), 993-1034.
[7]	Y. Feng and Y. Ying, Learning with correntropy-induced losses for regression with mixture of symmetric stable noise, Appl. Comput. Harmon. Anal., 48 (2020), 795-810. doi: 10.1016/j.acha.2019.09.001.
[8]	Z. C. Guo, T. Hu and L. Shi, Gradient descent for robust kernel-based regression, Inverse Probl., 34 (2018), Art. 065009. doi: 10.1088/1361-6420/aabe55.
[9]	R. He, W. S. Zheng and B. G. Hu, Maximum correntropy criterion for robust face recognition, IEEE Trans. Pattern Anal. Mach. Intell., 33 (2011), 1561-1576. doi: 10.1109/TPAMI.2010.220.
[10]	R. He, W. S. Zheng, B. G. Hu and X. W. Kong, A regularized correntropy framework for robust pattern recognition, Neural Comput., 23 (2011), 2074-2100. doi: 10.1162/NECO_a_00155.
[11]	P. W. Holland and R. E. Welsch, Robust regression using iteratively reweighted least-squares, Commun. Statist., 6 (1977), 813-827. doi: 10.1016/j.neucom.2016.12.029.
[12]	T. Hu, Q. Wu and D. X. Zhou, Distributed kernel gradient descent algorithm for minimum error entropy principle, Appl. Comput. Harmon. Anal., 49 (2020), 229-256. doi: 10.1016/j.acha.2019.01.002.
[13]	P. J. Huber, Robust Statistics., Wiley, New York, 2004.
[14]	J. Lin, L. Rosasco and D. X. Zhou, Iterative regularization for learning with convex loss functions, J. Mach. Learn. Res., 17 (2016), 2718-2755.
[15]	W. Liu, P. P. Pokharel and J. C. Principe, Correntropy: Properties and applications in non-gaussian signal processing, IEEE Trans. Signal Process., 55 (2007), 5286-5298. doi: 10.1109/TSP.2007.896065.
[16]	I. Pinelis et al., Optimum bounds for the distributions of martingales in banach spaces, Ann. Probab., 22 (1994), 1679-1706.
[17]	K. N. Plataniotis, D. Androutsos and A. N. Venetsanopoulos, Nonlinear filtering of non-gaussian noise, J. Intell. Robot. Syst., 19 (1997), 207-231. doi: 10.1023/A:1007974400149.
[18]	J. C. Principe, Information Theoretic Learning: Renyi's Entropy Entropy and Kernel Perspectives, Springer, New York, 2010. doi: 10.1007/978-1-4419-1570-2.
[19]	I. Santamaria, P. P. Pokharel and J. C. Principe, Generalized correlation function: Definition, properties, and application to blind equalization, IEEE Trans. Signal Process., 54 (2006), 2187-2197. doi: 10.1109/TSP.2006.872524.
[20]	S. Smale and D. X. Zhou, Estimating the approximation error in learning theory, Anal. Appl., 1 (2003), 17-41. doi: 10.1142/S0219530503000089.
[21]	S. Smale and D. X. Zhou, Learning theory estimates via integral operators and their approximations, Constr. Approx., 26 (2007), 153-172. doi: 10.1007/s00365-006-0659-y.
[22]	I. Steinwart, Oracle inequalities for support vector machines that are based on random entropy numbers, J. Complexity, 25 (2009), 437-454. doi: 10.1016/j.jco.2009.06.002.
[23]	I. Steinwart and A. Christmann, Support Vector Machines, Springer Science & Business Media, 2008.
[24]	X. Wang, Y. Jiang, M. Huang and H. Zhang, Robust variable selection with exponential squared loss, J. Amer. Statist. Assoc., 108 (2013), 632-643. doi: 10.1080/01621459.2013.766613.
[25]	B. Weng and K. E. Barner, Nonlinear system identification in impulsive environments, IEEE Trans. Signal Process., 53 (2005), 2588-2594. doi: 10.1109/TSP.2005.849213.
[26]	Q. Wu, Y. Ying and D. X. Zhou., Multi-kernel regularized classifiers, J. Complexity, 23 (2007), 108-134. doi: 10.1016/j.jco.2006.06.007.