独立分量分析可调速率相对梯度算法

Adjustable rate algorithm with relative gradient of ICA

摘要:

在独立分量分析的相对梯度算法中，要取得较好的效果，选取合适的学习速率是至关重要的。对于这个问题，文章提出了一种可调速率的相对梯度算法，随着迭代次数的变化，使相对梯度算法的学习速率作相应变化，从而较好地解决了收敛速度与稳定性的矛盾。在此基础上，将这个方法应用于盲信号分离并进行仿真，得到了满意的结果。可调速率相对梯度算法在独立分量分析中具有较好的前景。

Abstract:

In relative gradient algorithms of Independent Component Analysis(ICA), careful selection of step size is important to obtain good performance. In this study, an adjustable rate of relative gradient algorithm was proposed. With the changes of iteration number, the learning rate of relative gradient algorithm changed correspondingly，which solved the problem about the contradiction between the convergence rate and stability well. On this basis，this method was adopted in Blind Signal Separation(BSS) problems，and its effectiveness was validated by simulation. Adjustable rate algorithm with relative gradient has good prospects in independent component analysis.