Semi-supervised support vector machine (S3VM) is a generalization of support vector machine in semi-supervised learning. Compared with the support vector machine that needs to find the maximum margin dividing hyperplane, S3VM only considers the information of unlabeled samples and tries to find a dividing hyperplane that can separate the two types of labeled samples and pass through the low-density area of data.

S3VM Features

Based on the clustering hypothesis, S3VM attempts to standardize and adjust the decision boundary by exploring unlabeled data. In order to utilize unlabeled data, it is necessary to add two restrictions on unlabeled data points to the existing support vector machine "SVM".

Differences between S3VM and other support vector machines

TSVM

The transductive support vector machine "TSVM" and the semi-supervised support vector machine "S3VM" were proposed in the same year, and the main ideas of the algorithms and the optimization problems to be solved are similar, so the two concepts are interchangeable.

TSVM is mainly used for binary classification problems. It attempts to consider possible labeling indicators (Label Assignment) for unlabeled samples, that is, it tries to treat each unlabeled sample as a positive example or a negative example, and seeks a partitioning hyperplane with the maximum interval in the corresponding results.

Laplacian SVM

In addition to S3VM and TSVM, Laplacian SVM is also one of the relatively popular vector machines, which mainly explores the manifold structure of data through the Laplacian matrix of the graph.

These support vector machines have one thing in common, that is, they all directly estimate the unlabeled data categories.