为了解决当已分类完未标号样本,又有新的未标号样本的半监督学习问题,提出了能用于在线数据分类的半监督最接近支持向量机。在人工数据和CUI数据集上的实验显示,不因标号数据的增多而提高分类性能,未标号数据基本上不降低其分类性能,因此算法可在线使用。
To solve the problem of semi-supervised learning such that after the unlabeled data is labeled,new unlabeled data arrives, semi-supervised proximal support vector machine for on-line data classification is introduced.Experimental results on artificial and real data support that the performance of the proposed algorithm isn't improved as the number of labeled data increases and unlabeled data also doesn't decrease the performance.Thus the proposed algorithm can be used on-line.