一种基于FoBa算法思想的支持向量机稀疏SMO算法研究

A New Sparsity Smo Algorithm Research for SVM Based on FoBa

目前的支持向量机解析方法,如SMO算法在一定程度上解决传统支持向量机实现方法需要高额存储空间的问题,而对支持向量数目的约减并未过多关注,算法的稀疏性有待进一步提高。该文将FoBa算法对特征进行约减的思想引入SMO算法中,对训练产生的作用甚微的支持向量进行约减,提出了稀疏SMO算法。实验结果表明算法在提高预测速度上具有一定的竞争力。

SMO algorithm can settle the difficulty to a certain extent that traditional implement methed of SVM require a large number of storage and take no account of reducing the number of support vectors, the sparsity of algorrithm demand more improvement. This paper absorb the ideal of FoBa algorithm and reduce the number of support vectors that are not very important, bring forward a new algorithm named sparsity smo. We dernonstratc 1he efficiency of our approach by conducting experiments on benchmark database. Compared with smo, our algorithm have a better implement. Key Words SVM, SMO algorithm, sparsity, reducing support vectors, FoBa algorithm