Ensemble Clustering for Graphs

Poulin, Valérie; Théberge, François

doi:10.1007/978-3-030-05411-3_19

Cited by 17 publications

(29 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [14], we proposed ECG, a new graph clustering algorithm based on the concept of consensus clustering, and we compared it to other algorithms by re-creating the study in [11]. In this paper, we compared ECG with state-of-the-art algorihms over a wider range of graphs, showing ECG to be the best performing algorithm in most cases.…”

Section: Resultsmentioning

confidence: 99%

“…In a recent study [11], several state-of-the art algorithms implemented in the igraph [12] package were compared over a wide range of artificial networks generated via the LFR benchmark [13]. We recently introduced a new ensemble clustering algorithm for graphs (ECG), which compared favorably with leading algorithms from that study [14].The ECG algorithm is based on the concept of co-association consensus clustering. It is similar to other consensus clustering algorithms, in particular [15], but differs in two major points: (1) the choice of an algorithm that alleviates the resolution limit issue for the generation step, and (2) the restriction to endpoints of edges for co-occurrences of vertex pairs, which keeps low computational complexity.The rest of the paper is organized as follows.…”

mentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

Ensemble clustering for graphs: comparisons and applications

Poulin¹,

Théberge²

2019

Appl Netw Sci

Self Cite

View full text Add to dashboard Cite

We recently proposed a new ensemble clustering algorithm for graphs (ECG) based on the concept of consensus clustering. We validated our approach by replicating a study comparing graph clustering algorithms over benchmark graphs, showing that ECG outperforms the leading algorithms. In this paper, we extend our comparison by considering a wider range of parameters for the benchmark, generating graphs with different properties. We provide new experimental results showing that the ECG algorithm alleviates the well-known resolution limit issue, and that it leads to better stability of the partitions. We also illustrate how the ensemble obtained with ECG can be used to quantify the presence of community structure in the graph, and to zoom in on the sub-graph most closely associated with seed vertices. Finally, we illustrate further applications of ECG by comparing it to previous results for community detection on weighted graphs, and community-aware anomaly detection.Most networks that arise in nature exhibit complex structure [1, 2] with subsets of vertices densely interconnected relative to the rest of the network, which we call communities or clusters. Binary relational data-sets are typically represented as graphs G = (V, E), where vertices v ∈ V represent the entities, and edges e ∈ E represent the relations between pairs of entities. Graph clustering aims at finding a partition of the vertices V = C 1 ∪ . . . ∪ C l into good clusters. This is an ill-posed problem [3], as there is no universal definition of good clusters, leading to a wide variety of graph clustering algorithms [1, 4-10], with different objective functions. In a recent study [11], several state-of-the art algorithms implemented in the igraph [12] package were compared over a wide range of artificial networks generated via the LFR benchmark [13]. We recently introduced a new ensemble clustering algorithm for graphs (ECG), which compared favorably with leading algorithms from that study [14].The ECG algorithm is based on the concept of co-association consensus clustering. It is similar to other consensus clustering algorithms, in particular [15], but differs in two major points: (1) the choice of an algorithm that alleviates the resolution limit issue for the generation step, and (2) the restriction to endpoints of edges for co-occurrences of vertex pairs, which keeps low computational complexity.The rest of the paper is organized as follows. We briefly describe the ECG algorithm in Section 2, where we also recall some results from the previous comparison study. New results are included in the following three sections.In Section 3, we extend our study to a wider variety of graphs by varying the power law exponents of the LFR benchmark. Some of the advantages of ECG are its stability, and its ability to alleviate the well known resolution limit issue. We illustrate those properties in Section 4. We also take a closer look at the edge weights generated by the ECG algorithm, showing that they can be good indicators of the presence (or absence) of comm...

show abstract

Section: Resultsmentioning

confidence: 99%

mentioning

confidence: 99%

See 1 more Smart Citation

Ensemble clustering for graphs: comparisons and applications

Poulin¹,

Théberge²

2019

Appl Netw Sci

Self Cite

View full text Add to dashboard Cite

show abstract

“…, C . Note: In our implementation, we used the ensemble clustering algorithm for graphs (ECG) which is based on the Louvain algorithm and the concept of consensus clustering [10], and is shown to have good stability. We experiment with other algorithms in Section 4.…”

Section: Algorithmmentioning

confidence: 99%

An unsupervised framework for comparing graph embeddings

Kamiński

Prałat

Théberge³

2019

Journal of Complex Networks

Self Cite

View full text Add to dashboard Cite

Graph embedding is a transformation of vertices of a graph into set of vectors. Good embeddings should capture the graph topology, vertex-to-vertex relationship, and other relevant information about graphs, subgraphs, and vertices. If these objectives are achieved, they are meaningful, understandable, and compressed representations of networks. They also provide more options and tools for data scientists as machine learning on graphs is still quite limited. Finally, vector operations are simpler and faster than comparable operations on graphs.The main challenge is that one needs to make sure that embeddings well describe the properties of the graphs. In particular, the decision has to be made on the embedding dimensionality which highly impacts the quality of an embedding. As a result, selecting the best embedding is a challenging task and very often requires domain experts.In this paper, we propose a "divergence score" that can be assign to various embeddings to distinguish good ones from bad ones. This general framework provides a tool for an unsupervised graph embedding comparison. In order to achieve it, we needed to generalize the well-known Chung-Lu model to incorporate geometry which is interesting on its own rights. In order to test our framework, we did a number of experiments with synthetic networks as well as real-world networks, and various embedding algorithms. each vertex such that nearby vertices are more likely to share an edge than those far from each other. In a good embedding most of the network's edges can be predicted from the coordinates of the vertices. For example, in [9] protein interaction networks are embedded in low-dimension Euclidean space. Unfortunately, in the absence of a general-purpose representation for graphs, very often graph embedding requires domain experts to craft features or to use specialized feature selection algorithms. Having said that, there are some graph embedding algorithms that work without any prior or additional information other than graph structure. However, these are randomized algorithms that are usually not so stable; that is, the outcome is often drastically different despite the fact that all the algorithm parameters remain the same.Consider a graph G = (V, E) on n vertices, and several embeddings of its vertices to some multidimensional spaces (possibly in different dimensions). The main question we try to answer in this paper is: how do we evaluate these embeddings? Which one is the best and should be used? In order to answer these questions, we propose a general framework that assigns the divergence score to each embedding which, in an unsupervised learning fashion, distinguishes good from bad embeddings. In order to benchmark embeddings, we generalize well-known Chung-Lu random graph model to incorporate geometry. The model is interesting on its own and should be useful for many other problems and tools. In order to test our algorithm, we experiment with synthetic networks as well as real-world networks, and various embedding algorithms.The paper is...

show abstract

“…It can also be seen that the Louvain algorithm provides a relatively small number of predominantly large communities; this is a consequence of the resolution limit issue of modularity. To overcome this problem, it is possible to use the ECG approach (Ensemble Clustering for Graphs, Poulin and Théberge (2018)). Moreover, when many small communities exist in the network, there are approaches working better than the Louvain algorithm (e.g., InfoMap algorithm, Rosvall and Bergstrom (2008)).…”

Section: Zones In Community Structurementioning

confidence: 99%

Ego-zones: non-symmetric dependencies reveal network groups with large and dense overlaps

et al. 2019

View full text Add to dashboard Cite

The existence of groups of nodes with common characteristics and the relationships between these groups are important factors influencing the structures of social, technological, biological, and other networks. Uncovering such groups and the relationships between them is, therefore, necessary for understanding these structures. Groups can either be found by detection algorithms based solely on structural analysis or identified on the basis of more in-depth knowledge of the processes taking place in networks. In the first case, these are mainly algorithms detecting non-overlapping communities or communities with small overlaps. The latter case is about identifying ground-truth communities, also on the basis of characteristics other than only network structure. Recent research into ground-truth communities shows that in real-world networks, there are nested communities or communities with large and dense overlaps which we are not yet able to detect satisfactorily only on the basis of structural network properties.In our approach, we present a new perspective on the problem of group detection using only the structural properties of networks. Its main contribution is pointing out the existence of large and dense overlaps of detected groups. We use the non-symmetric structural similarity between pairs of nodes, which we refer to as dependency, to detect groups that we call zones. Unlike other approaches, we are able, thanks to non-symmetry, accurately to describe the prominent nodes in the zones which are responsible for large zone overlaps and the reasons why overlaps occur. The individual zones that are detected provide new information associated in particular with the non-symmetric relationships within the group and the roles that individual nodes play in the zone. From the perspective of global network structure, because of the non-symmetric node-to-node relationships, we explore new properties of real-world networks that describe the differences between various types of networks.

show abstract

Ensemble Clustering for Graphs

Cited by 17 publications

References 32 publications

Ensemble clustering for graphs: comparisons and applications

Ensemble clustering for graphs: comparisons and applications

An unsupervised framework for comparing graph embeddings

Ego-zones: non-symmetric dependencies reveal network groups with large and dense overlaps

Contact Info

Product

Resources

About