有效提高SVM参数搜索效率的样本集缩减策略

核函数及相关参数的选择是支持向量机中的一个重要问题, 它对模型的推广能力有很大的影响。当有大量样本参与训练的时候，寻找最优参数的网格搜索算法将消耗过长的时间。针对这一问题，提出一种舍弃非支持向量的样本点的策略，从而缩减了训练样本集。能够在基本保持原有测试准确度的前提下，将搜索时间减少一半。

Sample set shrinking strategy efficiently improving parameters seeking of support vector machines

The choice of kernel function and relative parameters plays an important role in Support Vector Machines (SVMs). It greatly influences the generalization performance of SVMs. It is time consuming to seek for optimal parameters when the training sample set is large. Concerning this problem, a sample set shrinking strategy was proposed. This method took some of the non-support-vector samples out of the training set; therefore efficiently reduced the set size. That is to say, with half the time consumed, a model can be constructed with testing accuracy just slightly changed.