作为当前最先进有效的密度估计算法,核密度估计(KDE)得到了广泛的研究。但是其二次的计算复杂度严重阻碍了KDE在具有海量高维数据的实际问题中的应用。为了排除算法计算性能上的障碍,研究者从不同角度提出了多种解决方案。在简要介绍KDE基本算法的基础上,简要分析了近年来提出的一些KDE的快速计算和逼近算法,以便为进一步的研究提供一定的支持与帮助。
As the most effective and powerful nonparametric density estimation technique,Kernel Density Estimation (KDE) has been widely analyzed.However,the practical use of KDE is not widespread because of high computational eomplexity due to the intensive use of the available data.In order to surmount this computational difficulty,researchers have proposed several schemes from different points of view.In this paper,after the brief introduction of conventional KDE,we analyze a few classes of fast KDE algorithms for the sake of further advance.