摘要
支持向量机(SVM)的文本分类算法被广泛应用,其中序列最小优化算法(SMO)是它的一个特例。SMO算法使用了块与分解技术,简单并且容易实现,但是它的收敛较慢,迭代次数较多。解决的办法是改进SMO算法中工作集的选择算法,并更新步长因子,目的是为了使目标函数尽可能地下降。文中基于这个目标提出了改进的SMO算法来进一步提高SVM的训练速度和分类的准确程度。
The support vector maehine（SVM） text classification algorithm is widely applied, and its special case is the sequence of minimum optimization（SMO） algorithm. SMO algorithm, with blocks and decomposition technology, is simple and easy to implement but slow in convergence, and has many iterative times. The solution for this is to improve the selection algorithm in the working set of SMO algorithm, and update the step factor, thus to make objective function decline as much as possible. With this goal, the improved SMO algorithm is proposed, thus to further improve the SVM training speed and the classification accuracy.
出处
《信息安全与通信保密》
2011年第12期63-64,67,共3页
Information Security and Communications Privacy
