摘要
研究了采用基于结构风险最小化原理的支持向量机对模式的分类方法 ,构造的分类模型结构简单 ,易于实现 ,且泛化能力明显提高 .该模型采用 2种核函数 ,分别以平面点集分类和手写字识别为例进行了仿真实验 .结果表明 ,将支持向量机用于模式识别不存在局部极小值问题 ,且不需进行网络迭代训练 ,求解速度明显高于前馈神经网络 .
This paper adopts supporting vector machine based on structure risk minimization principle in pattern recognition. Supporting vector machine is not only simpler in structure, but also better in performance, especially better in generalization ability. The model has been experimented on both classification of point set in plane and recognition of similar Chinese handwriting by two convolution of the inner product. Experiments exhitit that local minimums will be avoided in classification using support vector machine. The model is superior to neural network because it does not need a good deal of circulating operation.
出处
《大庆石油学院学报》
CAS
北大核心
2003年第2期59-61,共3页
Journal of Daqing Petroleum Institute
关键词
统计学习理论
支持向量机
机器学习
模式识别
statistical learning theory
support vector machine
machine learning
pattern recognition
