Semi-supervised Classification from Discriminative Random Walks

Abstract

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.

Supervised pattern classification based on optimum‐path forest

“…Besides, by changing the graph model and path-value function, one can derive other types of optimum-path forest classifiers, such as the unsupervised learning approach proposed in [23,24], which also relies on a different strategy to estimate prototypes. Most approaches for pattern classification based on graphs and/or paths in graphs are either unsupervised [25][26][27][28] or semi-supervised [29][30][31][32]. The proposed method can be easily extended to semi-supervised classification, given that the optimum-path forest can include unlabeled non-prototype samples.…”

Semi-supervised classification and betweenness computation on large, sparse, directed graphs

“…The same strategy can also be used in order to compute a betweenness measure. The second approach works on a trellis structure built from biased random walks on the graph, extending an idea introduced in [3]. These random walks allow to define a biased bounded betweenness for the nodes of interest, defined separately for each class.…”

Within-Network Classification Using Local Structure Similarity

“…Kernels on graphs Also related to our approach are classification methods based on graph kernels, including exponential and diffusion kernels [18], kernels using regularization operators [30], and kernels based on random walks [5,9,11,35]. However, these kernels tend to evaluate the affinity (proximity) between two nodes of a network indirectly based on the number and length of the paths between these nodes [9], thus subscribing to the homophily assumption.…”

“…The task of recovering the missing types of entities and links (i.e. node and edge labels) based on the available information, known as within-network classification, is a semi-supervised learning problem key to several applications like image processing [2,12], classifying document and web pages [5,6,21,32], classifying protein interaction and gene expression data [28], part-ofspeech tagging [19], detecting malicious or fraudulent activities [26], and recommending items to consummers [9,14].…”

“…While our method also uses similarity between nodes to define the class membership probabilities, it is more general in the sense that it allows the use of complex similarity measures that are not based on a vectorial representation of the neighborhood. Secondly, although methods based random walks have recently been proposed for the within-network classification problem [5,35], such methods strongly rely on the homophily assumption that nearby nodes are likely to have the same class. Following the success of structural kernels on the problem of graph classification [3,16,20], we present a novel relational classifier based on marginalized graph kernels [16], that evaluates the local structure similarity of two nodes in a network with parallel random walks.…”