增量支持向量机核函数的优化

支持向量机的核函数类型分为两类：局部核函数和全局核函数.局部核函数的值只受到相距很近数据点的影响，有很好的学习能力.全局核函数的值会受到距离较远数据点的影响，有很好的推广泛化能力.针对局部核函数学习能力良好但泛化能力差的缺点，提出一种结合局部核函数和全局核函数构造新联合函数的方法.实验结果表明，与局部核函数和全局核函数相比，新联合核函数有更好的预测能力，并且能够适应增量学习的过程.

Optimization of Kernel Function in Incremental Support Vector Machine

The Kernel function to support vector machines can be divided into two types: the local kernel function and the global kernel function. Because the local kernel function has excellent learning ability, but its generalization ability is limited, we structure a joint kernel function with two kinds of functions, so that it can combine the advantages of the two kinds of kernel functions. The experiment proves that the joint kernel function can adapt to the incremental learning process and it has better performance.