Semi-Supervised Classification From Discriminative Random Walks

This paper describes a novel technique, called D-walks, to tackle semi-supervised classification problems in large graphs. We introduce here a betweenness measure based on passage times during random walks of bounded lengths. Such walks are further constrained to start and end in nodes within the same class, defining a distinct betweenness for each class. Unlabeled nodes are classified according to the class showing the highest betweenness. Forward and backward recurrences are derived to efficiently compute the passage times. D-walks can deal with directed or undirected graphs with a linear time complexity with respect to the number of edges, the maximum walk length considered and the number of classes. Experiments on various real-life databases show the D-walks outperforms NetKit[5], the approach of Zhou and Scholkopf[15]and the regularized laplacian kernel[2]. The benefit of D-walks is particularly noticeable when few labeled nodes are available. The computation time of D-walks is also substantially lower in all cases.