A New Semidefinite Programming for Semi-supervised Support Vector Machines

摘要

The problem of semi-supervised SVMs, which constructs a SVM using both the training data and unlabeled data has been formulated as an integer optimization problem. In this paper, we present a semidefinite programming formation for the problem of semi-supervised SVMs by using the approach based on the convex relaxation. The aim is to use the efficient interior point algorithms to solve SDP model of the problem of semi-supervised SVMs and obtain an approximation of the optimal labeling.