Semi-supervised learning

Abstract

Semi-supervised learning is an important research topic in computational intelligence. In many real-world applications, such as text categorization, character recognition, natural language parsing, and so on, data labeling usually involves a lot of human efforts and therefore is expensive, while unlabeled data are relatively easy to obtain. For example, in web image classification, images can be easily obtained by a crawler. Semi-supervised learning studies how to utilize the unlabeled data to enhance the training process and to leverage the requirement on the amount of labeled data. Under the framework of semi-supervised learning, there are various learning methods motivated from different angles, including Expectation Maximization (EM) algorithms, Co-Training algorithms, graph-based learning algorithms and Semi-supervised Support Vector Machines (S3VMs). In this chapter, we will introduce the above mentioned semi-supervised learning algorithms. © 2013 Nova Science Publishers, Inc. All Rights Reserved.