By means of stochastic process theory, the bounded convergence of least mean square algorithm (LMS) is studied without data stationary assumption and ergodicity condition.

• 在无过程数据平稳性假设和各态遍历等条件下 ；运用随机过程理论研究了最小均方算法 (LMS)的有界收敛性 ；给出了估计误差的上界 ；论述了LMS算法收敛因子或步长的选择方法 ；以使参数估计误差上界最小 .