摘要
支持向量机是一种新的机器学习算法,与其它学习算法相比,它的最大优点是基于结构风险最小化原则,因而能够保证推广能力。推广能力估计是机器学习中的一个重要问题,是实现自适应调整、参数选择、模型选择的等方法的基础。本文详细比较当前较有影响的几种推广能力估计方法,指出了这些方法适应范围和优缺点,并结合各种方法的原理讨论了推广能力估计可能的发展方向。
Support Vector Machine (SVM) is a novel machine learning algorithm. Comparing with other algorithms, its distinct advantage is to utilize the Structure Risk Minimum rule as to guarantee its generalization performance. How to estimate generalization performance is an important problem in machine learning. Comparisons of some known algorithms are presented in the paper, advantages and disadvantages are analyzed and the way to further improvement is discussed.
出处
《电路与系统学报》
CSCD
2004年第4期86-91,96,共7页
Journal of Circuits and Systems
基金
"十.五"军事通信预研基金资助项目(4100104030)
