摘要
神经网络集成和支持向量机都是在机器学习领域很流行的方法。集成方法成功地提高了神经网络的稳健性和精度,其中选择性集成方法通过算法选择差异度大的个体,取得了很好的效果。而支持向量机更是克服了神经网络的局部最优,不稳定等缺点,也在多个方面取得了很好的结果。该文着重研究这两种方法在小样本多类数据集上的性能,在四个真实数据集上的结果表明,支持向量机性能要比神经网络集成稍好.
Both of neural network ensemble and support vector machines are popular methods in the machine learning community.The stability and accuracy of neural network can be significantly improved using ensemble techniques,in which the selective ensemble method can obtain excellent results by selecting the component which have more diversity.Support vector machines have performed excellent performance in many fields because of their generalization performance by realizing the principle of margin based structure risk minimization.This article will study on the performance of these approaches on multi-class data set,experiments on four real-world data sets show that support vector machines achieve better results than neural network ensemble does.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第1期46-47,119,共3页
Computer Engineering and Applications
基金
国家自然科学基金资助(编号:50174038)
关键词
神经网络集成
支持向量机
多值分类问题
nerual network ensemble,support vector machines,multiple classification problems