论文部分内容阅读
提出总间隔v-支持向量机(TM-v-SVM),该算法可取得比v-SVM更好的理论分类性能.研究表明TM-v-SVM等价于求解特征空间中的两个压缩凸包的最近点对.讨论压缩凸包的相关性质,并给出对应的几何算法.数值模拟实验表明TM-v-SVM和对应的几何算法可取得比其它算法更好的性能.
This paper proposes a generalized interval v-support vector machine (TM-v-SVM) which can achieve better theoretical classification performance than v-SVM.The results show that TM-v-SVM is equivalent to solving two compression convexities in feature space The nearest neighbor pairs of the package are discussed, and the related properties of the compressed convex hull are discussed and the corresponding geometric algorithms are given.The numerical simulations show that TM-v-SVM and the corresponding geometric algorithms can achieve better performance than other algorithms.