支撑矢量预选取的自适应投影算法  

Adaptive Projective Algorithm for Selecting Support Vector Beforehand

作　　者：丁爱玲[1] 刘芳[2] 曹伟[1] 

机构地区：[1]长安大学信息工程学院,西安710064 [2]西安电子科技大学计算机学院,西安710071

出　　处：《计算机工程与应用》2002年第19期116-118,共3页Computer Engineering and Applications

基　　金：国家自然科学基金(编号:60073053)

摘　　要：支撑矢量机是一种能在训练样本数很少的情况下达到很好分类推广能力的学习算法。但其在选择支撑矢量时却进行了大量不必要的运算,对此,该文提出了一种能够预选取支撑矢量的方法———自适应投影算法,该方法在不影响支撑矢量机的分类能力情况下,大大地减少了训练样本,提高了支撑矢量机的训练速度。仿真实验结果也验证了该方法的有效性和可行性。Support vector machine is a learning algorithm which has a good classification ability for limited training samples.But support vector machine processes a lot of unnecessary operation s when choosing the support vector.A new method called adaptive projective algorithm which is able to extract support vectors from given training examples is put forward for support vector algorithm.The method greatly reduces the training samples and the runtime of the improved program for support vector machine is shortened compared with the original one,while the classification ability of sup-port vector machine is unaffected.The experimental results show remarkable improvement of speed to support the idea.

关 键 词：支撑矢量预选取 自适应投影算法 支撑矢量机 模式识别 简化矢量 机器学习算法 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]