Traditional classifiers need labeled data to build a model. However, obtaining labeled instances
are often expensive, time consuming, and difficult while unlabeled instances are obtained relatively 
easy. But, there must be a way to use these unlabeled instances. Semi-supervised learning addresses 
this problem by using large amount of unlabeled data, together with labeled data to build a classifier. 
There are many semi-supervised learning methods which includes EM with generative mixture model, self-training, 
co-training, transductive support vector machine, and graph based methods. In graph based methods, instances 
are represented as a node in the graph and edges are represented as a similarity ( dissimilarity ) measure 
between two nodes. This report provides an overview of few graph based methods, which uses kernel function 
to determine the label of all unlabeled instances.