Kernel function clustering algorithm with optimized parameters

This paper deals with kernel function clustering algorithm with optimized parameter. Traditional clustering problems and solving algorithms are analyzed, and several limitations of traditional clustering algorithm are listed. These limitations are overcome by introducing kernel functions, which a nonlinear problem is transformed into a high dimension space. This paper proposes a kind of kernel function clustering algorithm with parameters optimized. Using these techniques, the nonlinear clustering problem in the high dimension space become simpler in which the inner distances of sample in the same class are shrunk and the distances between two class centers are increased relatively. The algorithm computing complexity is analyzed and a strategy of reducing complexity is presented. Also the primary factor of affecting clustering precision is discussed through an experiment example.