Improving semi-supervised support vector machines through unlabeled instances selection

摘要: Semi-supervised support vector machines (S3VMs) are a kind of popular approaches which try to improve learning performance by exploiting unlabeled data. Though S3VMs have been found helpful in many situations, they may degenerate and the resultant generalization ability be even worse than using labeled data only. In this paper, we reduce chance degeneration S3VMs. Our basic idea is that, rather all data, instances should selected such that only ones very likely exploited, while some highly risky avoided. We propose S3VM-us method hierarchical clustering select instances. Experiments on broad range sets over eighty-eight different settings show much smaller existing