Mixture correntropy for robust learning

Badong Chen, Xin Wang, Na Lu, Shiyuan Wang, Jiuwen Cao, Jing Qin

Research output: Journal article publicationJournal articleAcademic researchpeer-review

87 Citations (Scopus)


Correntropy is a local similarity measure defined in kernel space, hence can combat large outliers in robust signal processing and machine learning. So far, many robust learning algorithms have been developed under the maximum correntropy criterion (MCC), among which, a Gaussian kernel is generally used in correntropy. To further improve the learning performance, in this paper we propose the concept of mixture correntropy, which uses the mixture of two Gaussian functions as the kernel function. Some important properties of the mixture correntropy are presented. Applications of the maximum mixture correntropy criterion (MMCC) to extreme learning machine (ELM) and kernel adaptive filtering (KAF) for function approximation and data regression are also studied. Experimental results show that the learning algorithms under MMCC can perform very well and achieve better performance than the conventional MCC based algorithms as well as several other state-of-the-art algorithms.

Original languageEnglish
Pages (from-to)318-327
Number of pages10
JournalPattern Recognition
Publication statusPublished - Jul 2018


  • Correntropy
  • Extreme learning machine
  • Kernel adaptive filtering
  • Mixture correntropy
  • Robust learning

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence


Dive into the research topics of 'Mixture correntropy for robust learning'. Together they form a unique fingerprint.

Cite this