Active Learning with SVM

Abstract

With the increasing demand of multimedia information retrieval, such as image and video retrieval from the Web, there is a need to find ways to train a classifier when the training dataset is combined with a small number of labelled data and a large number of unlabeled one. Traditional supervised or unsupervised learning methods are not suited to solving such problems particularly when the problem is associated with data in a high-dimension space. In recent years, many methods have been proposed that can be broadly divided into two groups: semi-supervised and active learning (AL). Support Vector Machine (SVM) has been recognized as an efficient tool to deal with high-dimensionality problems, a number of researchers have proposed algorithms of Active Learning with SVM (ALSVM) since the turn of the Century. Considering their rapid development, we review, in this chapter, the state-of-the-art of ALSVM for solving classification problems.