摘要
针对经典支持向量机在增量学习中的不足,提出一种基于云模型的最接近支持向量机增量学习算法。该方法利用最接近支持向量机的快速学习能力生成初始分类超平面,并与k-近邻法对全部训练集进行约简,在得到的较小规模的精简集上构建云模型分类器直接进行分类判断。该算法模型简单,无须迭代求解,时间复杂度较小,有较好的抗噪性,能较好地体现新增样本的分布规律。仿真实验表明,本算法能够保持较好的分类精度和推广能力,运算速度较快。
Aiming at the limitations of incremental learning in classical SVM,this paper proposed an incremental PSVM（proximal SVM） learning algorithm based on cloud model.Employed the fast learning ability of PSVM to yield the initial classification hyperplane,and then,reduced all training datasets by using k-NN method and the plane.After that,utilized cloud model to directly discriminate analysis on the reduced dataset.The simple algorithm,with less computational time and better anti-noise ability,could well embody the distribution of incremental samples and could be solve without iteration.Experiment results show that the algorithm can not only keep well classification accuracy and generalization ability,but also improve the training speed.
出处
《计算机应用研究》
CSCD
北大核心
2011年第5期1685-1687,1691,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(60974126
60974050)
江苏省自然科学基金资助项目(BK2009094)
关键词
支持向量机
云模型
分类
增量学习
support vector machines（SVM）
cloud model
classification
incremental learning
