摘要
学习速率的优选问题是自适应ICA算法中的一个重要问题。本文建立了学习速率与相依性测度之间的一种非线性函数关系,以此为基础本文提出了一种新的变学习速率的自适应ICA算法。该算法具有初始阶段和未知系统时变阶段步长自动增大而稳态时步长很小的特点,克服了传统算法在稳态阶段步长调整过程中的不足,而且具有很快的收敛速度。计算机仿真结果与理论分析相一致,证实了该算法的性能。
An important point in adaptive ICA algorithm is the optimal learning rate. This paper establishes a nonlinear functional relationship between the learning rate and the dependent measure. On the basis of the functional relationship, the author presents the new algorithm of variable learning rate adaptive ICA. The learning rate of this algorithm increases automatically at the beginning of this algorithm or when unknown system is changing W tr ith time, it would be smaller during the steady state, so it overcomes the disadvantages of aditional algorithms in the process of step--size change of adaptive steady state, and has better convergence. The result of the computer simulations proves the algorithm performance.
出处
《工程地球物理学报》
2008年第4期493-498,共6页
Chinese Journal of Engineering Geophysics
基金
国家自然科学基金项目(编号:60672049)资助
关键词
独立成分分析
自然梯度
学习速率
independent component analysis （ICA）
natural gradient
learning rate
