Graph nodes clustering with the sigmoid commute-time kernel: A comparative study

Yen, Luh; Fouss, François; Decaestecker, Christine; Francq, Pascal; Saerens, Marco

doi:10.1016/j.datak.2008.10.006

Cited by 69 publications

(38 citation statements)

References 118 publications

(213 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Yet another interesting topic to be investigated is the clustering of large-scale graphs-we are currently working on extensions of graph kernel clustering applicable to large-scale graphs [62,63]. For a survey of community detection refer to [50].…”

Section: Resultsmentioning

confidence: 99%

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

Mantrach

Zeebroeck

Francq

et al. 2011

Pattern Recognition

Self Cite

View full text Add to dashboard Cite

a b s t r a c tThis work addresses graph-based semi-supervised classification and betweenness computation in large, sparse, networks (several millions of nodes). The objective of semi-supervised classification is to assign a label to unlabeled nodes using the whole topology of the graph and the labeling at our disposal. Two approaches are developed to avoid explicit computation of pairwise proximity between the nodes of the graph, which would be impractical for graphs containing millions of nodes. The first approach directly computes, for each class, the sum of the similarities between the nodes to classify and the labeled nodes of the class, as suggested initially in [1,2]. Along this approach, two algorithms exploiting different state-ofthe-art kernels on a graph are developed. 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. All the proposed algorithms have a linear computing time in the number of edges while providing good results, and hence are applicable to large sparse networks. They are empirically validated on medium-size standard data sets and are shown to be competitive with state-of-the-art techniques. Finally, we processed a novel data set, which is made available for benchmarking, for multi-class classification in a large network: the U.S. patents citation network containing 3M nodes (of six different classes) and 38M edges. The three proposed algorithms achieve competitive results (around 85% classification rate) on this large network-they classify the unlabeled nodes within a few minutes on a standard workstation.

show abstract

Section: Resultsmentioning

confidence: 99%

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

Mantrach

Zeebroeck

Francq

et al. 2011

Pattern Recognition

Self Cite

View full text Add to dashboard Cite

show abstract

“…Further work will be devoted to the extension and to the study of other centrality measures that lie in our proposed framework as well as the development of new clustering algorithms dealing, for instance, with other dissimilarity measures (for instance weighted distance, parametric Pearson product moment correlation, or angular distance (Yen et al, 2009)). We will also investigate the possibility of adapting multiple contexts (layers).…”

Section: Discussionmentioning

confidence: 99%

“…As Yen et al (2009) mentioned, spectral techniques have been applied in a wide variety of contexts including high performance computing, image segmentation, web pages ranking, information retrieval, data clustering, and dimensionality reduction.…”

Section: Spectral Clustering Approachesmentioning

confidence: 99%

A novel framework in complex network analysis: Considering both structure of relations and individual characteristics in closeness centrality computation

Barzinpour¹,

Ahmadi²

2013

ijiec

View full text Add to dashboard Cite

In this paper, we develop a novel framework for defining radial measures of centrality in complex networks. This framework is based on the combination of two approaches: social network analysis and traditional social science approach by considering both structure of relations and individual characteristics. It is always an important issue to detect communities in complex networks as efficiently as possible to understand both the structure and function of the networks and to interpret radial centrality measures. Therefore, we propose spectral clustering by determining the best number of communities as a prerequisite stage before finding radial measures. Based on the proposed framework, an algorithm to compute the closeness centrality in complex networks is developed. We test the proposed algorithm on Zachary's karate club network, which is considerably used as a benchmark for community detection in a network. The preliminary results indicate that the new method is efficient at detecting both good inter-cluster closeness centrality and the appropriate number of clusters.

show abstract

“…With these eight benchmarks, we compared MVSim with: Cosine, LSA, CTK [10] and χ-Sim [6] that are five classical similarity or co-similarity measures; ITCC [4] a well-known co-clustering system; MVSC [7] a multi-view algorithm. Finally, we ran two basic versions of MVSim without iteration (no feedback loop nor damping factor), to verify that our results are significantly better than those obtain by simply averaging the similarity matrices computed from each R i,j ; we tested two similarity measures : cosine (Merge Cosine) and χ-Sim (Merge χ-Sim).…”

Section: Methodsmentioning

confidence: 99%

An Architecture to Efficiently Learn Co-Similarities from Multi-view Datasets

Bisson

Grimal

2012

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. In this paper, we introduce the MVSim architecture which is able to cluster multi-view datasets (i.e. datasets containing several objects linked together by different relations), by using several instances of a co-similarity algorithm. We show that this framework provides better results than existing approaches, while reducing both time and space complexities thanks to an efficient parallelization of the computations. This approach allows to split large datasets into a set of smaller ones.

show abstract

Graph nodes clustering with the sigmoid commute-time kernel: A comparative study

Cited by 69 publications

References 118 publications

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

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

A novel framework in complex network analysis: Considering both structure of relations and individual characteristics in closeness centrality computation

An Architecture to Efficiently Learn Co-Similarities from Multi-view Datasets

Contact Info

Product

Resources

About