论文部分内容阅读
提出了一种基于核密度估计(KDE)的核分类算法SKVM。SKVM构建了可稀疏化分类器参数的模型,并根据两类样本的密度差信息构造了约束函数,使用标准二次规划方法求解分类器参数。其优势在于能在实现有效分类的同时降低决策函数的计算复杂度;而且通过调节SKVM的窗宽值可以控制分类器参数的稀疏率,实验结果表明了上述优势。
A kernel classification algorithm SKVM based on kernel density estimation (KDE) is proposed. SKVM constructs a model of sparse classifier parameters and constructs a constraint function based on the density difference information of the two types of samples. The standard quadratic programming method is used to solve the classifier parameters. The advantage of this method is that it can reduce the computational complexity of the decision function while achieving effective classification. Moreover, the sparse rate of the classifier parameters can be controlled by adjusting the window width of SKVM. The experimental results show the above advantages.