随机梯度估值在盲均衡算法中的影响

Influence of Stochastic Gradient Estimation on Blind Equalization Algorithm

利用代价函数求梯度值,再利用梯度值进行迭代是一种最速下降法,在各种盲均衡算法中广泛使用,其收敛速度较快,收敛误差较小,但未考虑信道中随机多变的噪声对其产生的影响.该文引入随机梯度估值,在梯度向量中加入噪声成分,结合梯度估值引起的权偏差相关系数,分析其对盲均衡算法的影响.理论研究和仿真结果表明,随机梯度估值引起权值偏差,影响着算法的收敛速度和剩余误差.

This paper presents the stochastic gradient estimation, which adds the noise in the channel to the gradient. Based on the correlation coefficient, an analysis is made of the effect of the stochastic gradient estimation on the blind equalization algorithm. Theoretical research shows that the stochastic gradient estimation lead to the deviation of the weight, which determines the size of the residual error.