The contradiction of variable step size least mean square(LMS)algorithm between fast convergence speed and small steady-state error has always existed.So,a new algorithm based on the combination of logarithmic and sym...The contradiction of variable step size least mean square(LMS)algorithm between fast convergence speed and small steady-state error has always existed.So,a new algorithm based on the combination of logarithmic and symbolic function and step size factor is proposed.It establishes a new updating method of step factor that is related to step factor and error signal.This work makes an analysis from 3 aspects:theoretical analysis,theoretical verification and specific experiments.The experimental results show that the proposed algorithm is superior to other variable step size algorithms in convergence speed and steady-state error.展开更多
We provide some lower bounds on the deficit in the Gaussian logarithmic Sobolev inequality in terms of the so-called Stein characterization of the Gaussian distribution.The techniques are based on the representation o...We provide some lower bounds on the deficit in the Gaussian logarithmic Sobolev inequality in terms of the so-called Stein characterization of the Gaussian distribution.The techniques are based on the representation of the relative Fisher information along the Ornstein-Uhlenbeck semigroup by the Minimum Mean-Square Error from information theory.展开更多
基金the National Natural Science Foundation of China(No.51575328,61503232).
文摘The contradiction of variable step size least mean square(LMS)algorithm between fast convergence speed and small steady-state error has always existed.So,a new algorithm based on the combination of logarithmic and symbolic function and step size factor is proposed.It establishes a new updating method of step factor that is related to step factor and error signal.This work makes an analysis from 3 aspects:theoretical analysis,theoretical verification and specific experiments.The experimental results show that the proposed algorithm is superior to other variable step size algorithms in convergence speed and steady-state error.
基金Grants No.F1R-MTH-PUL-15CONF and No. F1R-MTH-PUL-15STAR at Luxembourg University
文摘We provide some lower bounds on the deficit in the Gaussian logarithmic Sobolev inequality in terms of the so-called Stein characterization of the Gaussian distribution.The techniques are based on the representation of the relative Fisher information along the Ornstein-Uhlenbeck semigroup by the Minimum Mean-Square Error from information theory.
