摘要
在各种梯度类自适应算法中,如何决定递推步长μ是一个很关键的问题。但迄今为止,无论是采用固定步长还是一些变步长方法,这个问题都没有很好地解决。本文基于改进的梯度技术,提出一种决定每步递推的最佳步长μ_(pt)的算法。模拟结果表明,无论信号环境如何,最佳步长算法可以大大地提高算法的收敛速度,也解决了实际环境下的步长选择问题。
In all kinds of adaptive algorithm based on gradient, how to choosethe step size is a very important problem.But till now, despite of using con-stant step or variable step algorithms, the problem hasn't been solved well.Based on the improved gradient technique, this paper brings up an optimal stepalgorithm.It is indicated by analogous results that the convergent rate can behighly improved despite of the signal circumstances. The problem of choosingthe step size in real case is also solved.
出处
《信号处理》
CSCD
北大核心
1989年第4期234-239,共6页
Journal of Signal Processing
