【摘要】 文章提出了两种快速分类的方法——基于最小超球体的平分最近点法和基于最小超球体的按比例划分法。前者只对分别包含正、负类训练点的两类超球体线性可分的情形有效,后者则适用于线性可分和近似线性可分的两类分类问题,且在确定分划超平面时融入了对训练集分布特征的考虑。两种方法皆借鉴了平分最近点法的思想,结合超球体的几何特征,用解析几何方法就可求得分划超平面,从而避免了求解二次规划,大大缩短了训练时间,减小了内存占用量,尤其在处理大规模数据集时优势更为明显。两种方法的特点及其和平分最近点法的对比在实证中都给予了分析说明。更多还原

【Abstract】 Two fast algorithms for classification are presented in this paper--halving the nearest points and dividing the nearest points proportionally based on the minimal hyper-sphere.There are something different in their applicative ranges.All of them referred to the idea of halving the nearest points method and the geometry character of hyper-sphere,and we can avoid to solve the traditional quadratic program.Compared with halving the nearest points method,the learning algorithm not only increases the training speed,but also decreases the consumption of EMS memory.The characters of the two fast algorithms and their comparison with halving the nearest points method are shown in part 4.更多还原