支持向量机的若干新进展

Some New Developments on Support Vector Machine

支持向量机是九十年代中期发展起来的机器学习技术 ,与传统的人工神经网络不同 ,前者基于结构风险最小化原理 ,后者基于经验风险最小化原理 .实验表明 ,支持向量机不仅结构简单 ,而且技术性能尤其是泛化能力明显提高 .本文是一篇综述 ,介绍支持向量机研究的一些新进展 。

Support vector machine is new machine learning technique developed from the middle of 1990s. Being different from traditional neural network, it is based on structure risk minimization principle, while the latter on empirical risk minimization principle. A large number of experiments show that, comparing with traditional neural network, support vector machine has not only simpler structure, but also better performances, especially better generalization ability. Some new developments on support vector machine are introduced so as to draw our attention.