摘要
介绍了概率波函数、薛定锷方程、量子势能和量子聚类(Quantum clustering,QC)算法。给出了量子势能确定聚类中心的量子力学依据,强调其重点在于聚类中心的确定而非聚类边界;指出了QC算法中的波函数相当于一个核函数,其中的尺度参数的实质是核宽度调节参数,并给出一种有关它的直方图估算方法。基于此,本文提出一种基于核宽度调节参数估算的改进的量子聚类(Parameter-estimated quantum clustering,PeQC)算法,可克服QC算法中要通过多次训练来最后选取参数的不足。通过与QC算法的实验比较,证明该算法有较高的聚类效能,比标准量子聚类算法简单可行、易操作。
Probability wave function,Schrdinger equation,quantum potential and quantum clustering（QC） algorithm are introduced.By analyzing the physical essence of the quantum clustering algorithm,the quantum potential is based on the theoretical basis of QC algorithm in quantum mechanics,thus it is used to determine the cluster centers rather than the clustering boundary.Moreover,the probability wave function of QC algorithm is a kernel function for converting the nonlinear input space to a Hilbert space,and the scale-parameter δ therein is accordingly revealed to be the corresponding kernel width.Based on the above,the parameter-estimated quantum clustering algorithm PeQC is proposed to overcome the defect,that is,the parameter is often needed to be estimated by experiments many times.Experimental results demonstrate that the algorithm can be easily realized than QC algorithm.
出处
《数据采集与处理》
CSCD
北大核心
2008年第2期211-214,共4页
Journal of Data Acquisition and Processing
关键词
量子聚类算法
核宽度调节参数
量子势能
参数估计
quantum clustering algorithm
kernel width adjustive parameter
quantum potential
parameter estimation
