基于微分竞争学习的参数估计器

A New Parameter Estimator Using Differential Competitive Learning

推导了基于微分竞争学习的参数估计器（ＤＣＬＰＥ），为解决没有统计先验信息、线性或非线性观测过程以及待估参数属于多类统计模式的参数估计问题提供了一种有效方法．

When there is no prior statistical information of the observation noise, and when the observation process is nonlinear and when the parameters to be determined belong to different patterns in the mixed observation process, then, to the authors best knowledge, existing algorithms in the open literature can not locate global optimum of the performance functional of parameter estimation. In this paper the authors propose a new algorithm or parameter estimator that can locate such global optimum and obtain the optimal parameter estimate. This new algorithm is called Differential Competitive Learning Parameter Estimator (DCLPE) which is derived from DCL paradigm. DCLPE is capable of searching globally because it uses a parallel form of learning and multi-agent search strategy.