摘要
研究了支持向量机与自组织神经网络的原理,利用支持向量机的小样本学习与推广能力强的特点,结合自组织神经网络良好的学习能力与收敛速度,实现了对支持向量机算法的改进。利用Lincoln实验室入侵检测系统评估数据集合对改进算法进行测试,并将实验结果与BP神经网络进行了比较,结果表明,改进的算法在检测精度与训练时间方面均优于BP神经网络。
This paper studies the principle of the support vector machine and self-orgnizing neural network.An improved algorithm has been worked out using small samples and strong generalization ability of support vector machine combined with good training ability and convergence rate of self-orgnizing neural network.The improved algorithm was tested using Lincoln laboratory intrusion detection system evaluation data set.The experiment results of the improved algorithm and BP neural network were compared.The result shows that the improved algorithm is better than BP neural network in detection precision and training time.
出处
《科技通报》
北大核心
2012年第10期158-159,162,共3页
Bulletin of Science and Technology
基金
2011年黑龙江省高等教育教学改革工程项目的部分研究成果
关键词
支持向量机
最优超平面
自组织神经网络
非线性变换
support vector machine
optimal hyperplane
self-orgnizing neural network
nonlinear transform