论文部分内容阅读
在一定条件下,基于最小累积平方误差(ISE)准则的高斯核密度估计与最小包含球(MEB)等价.在此基础上提出了一种含团状隐私数据保护的MEB学习方法,称为隐私团校准的MEB(PCC-MEB)方法;同时,通过引入模糊隶属度函数将PCC-MEB拓展为模糊的PCC-MEB(FPCC-MEB),从而解决二类及多类问题中区域不可分问题.人造和真实数据集上的实验结果表明,所提出方法具有较好的性能.
Under certain conditions, the Gaussian kernel density estimation based on the minimum cumulative squared error (ISE) criterion is equivalent to the minimum inclusion sphere (MEB) .An MEB learning method with clumped privacy data protection (PCC-MEB) method for privacy group calibration. At the same time, PCC-MEB is extended to fuzzy PCC-MEB (FPCC-MEB) by introducing fuzzy membership function, so as to solve the indivisible problem of the two kinds and many kinds of problems. Experimental results on artificial and real data sets show that the proposed method has better performance.