基于边界支持向量的自适应增量支持向量机
Adaptive Incremental Support Vector Machine Based on Border Support Vectors
-
摘要: 该文提出一种基于边界支持向量的自适应增量支持向量机，对每轮训练的样本集提取其边界支持向量，从而减少训练向量数目，提高训练效率。通过自适应调整参数，可以更好地适应新增样本。采用 UCI（University of California Irvine）机器学习数据库和Statlog数据库对本文方法进行验证，实验结果表明本文方法的训练时间优于标准支持向量机和一般增量支持向量机。其分类精度也明显优于一般增量支持向量机，在训练数据较少时，其分类精度与标准支持向量机相差不大，但随着训练数据的增加，分类精度逐渐超越标准支持向量机。该文的方法更适合大规模数据集的增量学习。Abstract: An adaptive incremental support vector machine based on border support vectors has been pro‐posed ,in each training set ,the border support vectors have been extracted in order to reduce the number of training samples and increase training efficiency .By adaptive adjustment ,the parameter can adapt to the increased samples .The UCI datasets and Statlog datasets have been used to test the proposed algorithm . Experiment results show the approach is significantly better than the standard support vector machine and incremental support vector machine in training time and it outperforms incremental support vector machine in classification accuracy .If the number of training samples is small ,the accuracy is similar with standard SVM ,while with the increasing of training samples ,the accuracy was beyond the standard SVM gradual‐ly ,it supports that the method is suitable for increasing learning for large‐scare datasets .
-
-
计量
- 文章访问数: 355
- HTML全文浏览数: 156
- PDF下载数: 0
- 施引文献: 0
下载: