摘要
提出一种处理样本中含有未确知信息(一种不确定性信息)的支持向量机—未确知支持向量机(Unascertained support vector machine,USVM)算法.首先,以未确知数学为基础,将含有未确知信息的分类问题转化为求解未确知机会约束规划问题.然后,将其转化为与其等价的二次规划.据此给出未确知支持向量机.理论分析和试验结果均表明,该算法是有效、可行的.
In this paper, we propose an unascertained support vector machine （USVM） for classification problem with unascertained information （a kind of uncertain information）. First, based on unascertained mathematics, classification problems containing unascertained information are converted to solving unascertained chance constrained programming problems. Then, the unascertained chance constrained programming is converted to its equivalent quadratic programming. Based on these theories, the algorithm of a unascertained support vector machine is given. Theoretical analysis and experiments show that the proposed USVM is effective and feasible.
出处
《自动化学报》
EI
CSCD
北大核心
2013年第6期895-901,共7页
Acta Automatica Sinica
基金
国家自然科学基金(10926198
11201426)
浙江省自然科学基金(LQ12A01020)资助~~
关键词
机器学习
未确知支持向量机
未确知信息
未确知数
Machine learning, unascertained support vector machine （USVM）, unascertained information, unascertained number
