一种变步长CMA算法的研究

Research on an improved CMA algorithm with variable step size

采用固定步长的CMA算法使得收敛速度与精度存在制约关系,导致收敛效果不理想。为解决此问题,提出了一种变步长的CMA算法,即在算法收敛刚开始时增大步长值,从而加快收敛速度,在算法即将到达收敛点时,减小步长值,从而降低稳态误差。算法分析和仿真结果表明,改进的算法具有更快的收敛速度。

Using CMA algorithm with fixed step size, the convergence speed and accuracy are restricted, resulting in no convergence effect. In order to solve this problem, a variable step size CMA algorithm is proposed, which increases the step size at the beginning of convergence to accelerate the convergence speed, and decreases the step size when approaching the convergence point to reduce the steady-state error. The results of algorithm analysis and simulation show that the improved algorithm has a faster convergence rate.