Fast Multi-Scale Detection of Relevant Communities in Large-Scale Networks

Abstract

Nowadays, networks are almost ubiquitous. In the past decade, community detection received an increasing interest as a way to uncover the structure of networks by grouping nodes into communities more densely connected internally than externally. Yet most of the effective methods available do not consider the potential levels of organisation, or scales, a network may encompass and are therefore limited. In this paper we present a method compatible with global and local criteria that enables fast multi-scale community detection on large networks. The method is derived in two algorithms, one for each type of criterion, and implemented with 6 known criteria. Uncovering communities at various scales is a computationally expensive task. Therefore this work puts a strong emphasis on the reduction of computational complexity. Some heuristics are introduced for speed-up purposes. Experiments demonstrate the efficiency and accuracy of our method with respect to each algorithm and criterion by testing them against large generated multi-scale networks. This study also offers a comparison between criteria and between the global and local approaches. In particular our results suggest that global criteria seem to be more robust to noise and thus more accurate than local criteria.

Collective Behaviour in Video Viewing: A Thermodynamic Analysis of Gaze Position

“…Since we are interested in whether there are groups of viewing behaviours, we analyse the network structure of these interactions J ij . We use a conventional clustering technique [28] to group together subjects with similar entries in the interaction matrix J ij , as shown in Fig 4. Separate communities can be seen in the groups of blocks along the diagonal; each group is a different community, and these are the same groups seen in the snapshot in Fig 1B.…”

“…It is not surprising that strong average C ij leads to clear modular networks in interactions J ij , as those videos with strong average correlations also displayed strong average interactions. However, the deeper community structure [28, 31] found in the interactions is not immediately apparent in pairwise correlations of the gaze direction (see Fig 6). This is likely due to the presence of “noise” in the raw data such as an apparent correlation C ij between two subjects which only arise from their each being similar to a third subject; these two subjects would not have a strong interaction J ij , leading to differences in the community structure of C ij and J ij for the same video.…”

“…Each row and column indicates a viewer, and viewers have been sorted to group rows/columns with similar entries together using a conventional clustering algorithm [28]. Self-correlations are set to zero for clarity.…”

Community mining with new node similarity by incorporating both global and local topological knowledge in a constrained random walk

“…But, the threshold may not be stationary because the community structure in nature is very complicated such that there could be multiple values but not one suitable value. This will be interesting since by varying the parameter α, we are possible to reveal the multi-level community structures at different scales in the network [53][54][55][56].…”

A scalable approach for Contradiction Detection driven by Opinion mining

“…The community structure of opinion holders is identified based on the opinion similarity and the social features of holders. The approach uses the Infomap community detection algorithm [12] and the Multi-Scale Community Detection framework [9]. The communities' structure helps in the following ways for detecting contradictions: (1) for a given holder, by contrasting a new opinion to his/her community's opinion on the same target, (2) contrasting a given opinion with the opinion of the holder's social community.…”