摘要
支持向量机是由Vapnik等提出的建立在统计学习理论基础上的一种新的机器学习方法,由于其使用结构风险最小化原则代替经验风险最小化原则,又由于其应用了核函数思想,它可以较好地解决非线性问题;人工神经网络(ANN)已经较成功解决模式识别和任意非线性函数回归问题,但是存在训练样本不足,并可能出现过拟合现象。SVM的结构风险最小化算法引起了科学界的关注,对传统基于经验风险最小化的神经网络算法提出了挑战,文章介绍了SVM和ANN的基本原理,并对二者在巢湖富营养化水平评价上做对比研究,结果表明,ANN比较容易陷入局部最优,支持向量机评价结果更加符合实际。
Support Vector Machine（SVM）,a new statistic-learning-theory based method proposed by V.Vapnik,can be used to tackle non-linear problems owing to its kernel function,i.e.,structural risk minimization principle instead of the experimental risk minimization.Artificial neural network（ANN） has been applied for years to resolving the problems of pattern recognition and regression,but some questions still remained unresolved.Based on the water quality data from 2001 to 2003 of Chaohu Lake,with which evaluations of the Lake＇s eutrophication level were performed by use of both SVM and ANN,this paper gave a result of this comparative study that indicated that SVM was more practical in describing the lake＇s eutrophication status.
出处
《环境科学与技术》
CAS
CSCD
北大核心
2012年第1期173-177,共5页
Environmental Science & Technology
关键词
支持向量机
神经网络
核函数
富营养化
support vector machine（SVM）
artificial neural network
kernel function
eutrophication
