Variable Kernel Width Algorithm of Generalized Maximum Correntropy Criteria for Censored Regression
Haiquan Zhao, Bing Chen, Yingying Zhu, Xiaoqiong He, Zeliang Shu
Abstract
The constant kernel width of generalized maximum correntropy criteria (GMCC) has arisen that the steady-state error and convergence speed can be mutually exclusive. To solve this problem, this brief proposes the variable kernel width (VKW) GMCC algorithm. Actually, due to the censored problem, the output data value beyond the limit of the recording device can not be well observed. In this case, we further developed a variable kernel width GMCC algorithm based on censored regression (CR-VKWGMCC). Simulation results show that the proposed CR-VKWGMCC algorithm has excellent performance in both Gaussian and non Gaussian noise.
Topics & Concepts
Kernel (algebra)MathematicsVariable (mathematics)AlgorithmGaussianConvergence (economics)Kernel regressionRegressionComputer scienceMathematical optimizationApplied mathematicsStatisticsCombinatoricsMathematical analysisPhysicsEconomic growthEconomicsQuantum mechanicsAdvanced Adaptive Filtering TechniquesSpeech and Audio ProcessingBlind Source Separation Techniques