一种改进的FastICA算法

An Improved FastICA Algorithm

FastICA算法是一种快速独立分量分析(Independent Component Analysis:ICA)算法,但它是基于牛顿迭代方法和合理近似的一种算法,所以具有改进空间.近年来提出了许多改进的具有更高阶收敛性质的牛顿迭代方法.将一种3阶收敛的牛顿迭代方法引入ICA算法的推导中,在合理近似的基础上,提出了一种改进的两步迭代FastICA算法.与传统FastICA算法相比,提出的改进的FastICA算法一次迭代的计算量有所增加.但是,实验结果表明,新提出的改进的FastICA算法更稳健、具有更快的收敛速度.

FastICA algorithm is a kind of Independent Component Analysis(ICA) algorithm,but it is an algorithm based on Newton iterative method and reasonable approximation,so it has room for improvement.In recent yeaxs,many improved Newton iterative methods with higher order convergence properties have been proposed.In this paper,a Newton iterative method with 3 rd order convergence is introduced into the derivation of ICA algorithm.Based on reasonable approximation,an improved two-step iterative FastICA algorithm is proposed.Compared with the traditional FastICA algorithm,the improved FastICA algorithm proposed in this paper increases the computational complexity of one iteration.However,the experimental results show that the proposed improved FastICA algorithm is more robust and has a faster convergence rate.