期刊论文

中文

仲恺农业工程学院学报

1674-5663

2007

20

2

31-35

10.3969/j.issn.1674-5663.2007.02.007

本校为其他机构

提出了一种基于β因子历史样本淘汰机制的在线学习算法.对UCI标准数据集中的部分样本集的测试结果表明:该机制有效地淘汰了一些样本,在保持了分类精度和泛化能力的情况下,大大加快了增量学习的训练速度.

A new online learning algorithm was proposed in this paper, which was based on the β factor of dis-carding history sample. Several standard benchmark data sets in the UCI were tested through the new online learning algorithm. The results showed that some samples could be discarded effectively through β factor, which leaded to fast incremental learning with no influenc...