Given an arbitrary group of computers, how to identify abnormal changes in their communication pattern? How to assess if the absence of some communications is normal or due to a failure? How to distinguish local from global events when communication data are extremely sparse and volatile? Existing approaches for anomaly detection in interaction streams, focusing on edge, nodes or graphs, lack flexibility to monitor arbitrary communication topologies. Moreover, they rely on structural features that are not adapted to highly sparse settings. In this work, we introduce MAD, a novel Multi-scale Anomaly Detection algorithm that (i) allows to query for the normality/abnormality state of an arbitrary group of observed/non-observed communications at a given time; and (ii) handles the highly sparse and uncertain nature of interaction data through a scoring method that is based on a novel probabilistic and multi-scale analysis of sub-graphs. In particular, MAD is (a) flexible: it can assess if any time-stamped subgraph is anomalous, making edge, node and graph anomalies particular instances; (b) interpretable: its multi-scale analysis allows to characterize the scope and nature of the anomalies; (c) efficient: given historical data of length N and M observed/non-observed communications to analyze, MAD produces an anomaly score in O (NM); and (d) effective: it significantly outperforms state-of-the-art alternatives tailored for edge, node or graph anomalies.

Graph autoencoders (GAE), also known as graph embedding methods, learn latent representations of the nodes of a graph in a low-dimensional space where the structural information is preserved. While real-world graphs are generally dynamic, only a few embedding methods handle the temporal dimension: Even though they have proven their reliability, the majority of the embedding techniques address the case of static networks and present poor performances when applied to temporal ones.

Community detection emerged as an important exploratory task in complex networks analysis across many scientific domains. Many methods have been proposed to solve this problem, each one with its own mechanism and sometimes with a different notion of community. In this article, we bring most common methods in the literature together in a comparative approach and reveal their performances in both real-world networks and synthetic networks. Surprisingly, many of those methods discovered better communities than the declared ground-truth communities in terms of some topological goodness features, even on benchmarking networks with built-in communities. We illustrate different structural characteristics that these methods could identify in order to support users to choose an appropriate method according to their specific requirements on different structural qualities.

In epistemic communities, people are said to be selected on their contribution in knowledge to the project (articles, codes, etc.). However, the socialization process is an important factor for inclusion, sustainability as a contributor, and promotion. Finally, what matters for being promoted? Being a good contributor? Being a good animator? Knowing the boss? We explore this question by looking at the election process for administrators in the English Wikipedia. We used the candidates' revisions and/or social attributes to construct a predictive model of promotion success, based on the candidates' past behavior and a random forest algorithm. Our model explains 78% of the results, which is better than the former models. It also helps to refine the explanation of the election process.

Epistemic communities are said to be project-oriented communities of experts, evaluated on their contribution in terms of knowledge, where the main criterion for promotion is knowledge production [3]. However, [5], for Wikipedia, [7], for open source, have argued that taking responsibility is an additional step from being a regular contributor, and social interactions with peers may be an additional requirement for being promoted [6].

Attributed graphs can be described using two dimensions: first a structural dimension that contains the social graph, e.g. the actors and the relationships between them, and second a compositional dimension describing the actors, e.g. their profile, their textual publications, the metadata of the videos they share, etc. Each of these dimensions can be used to explain different phenomena occurring on the social network, whether from a connectivity or an thematic perspective. This paper claims that the integration of both dimensions would allow researchers to analyze real social networks from different perspectives. We present here a novel approach to the community detection problem with the integration of the two dimensions composing an attributed graph. We show how to integrate but also how to control the integration of two different partitions, one based on the links, the other one based on the attributes. The resulting partition exhibits interesting properties, such as dense and homogeneous groups of actors, revealing new types of communities to the analyst. Because we use a contingency matrix, and because the analyst may invent new ways of combining rows and columns, we open new perspectives for the exploration of attributed social networks.

Due to the explosion of social networking and the information sharing among their users, the interest in analyzing social networks has increased over the recent years. Two general interests in this kind of studies are community detection and visualization. In the first case, most of the classic algorithms for community detection use only the structural information to identify groups, that is, how clusters are formed according to the topology of the relationships. However, these methods do not take into account any semantic information which could guide the clustering process, and which may add elements to conduct further analyses. In the second case most of the layout algorithms for clustered graphs have been designed to differentiate the groups within the graph, but they are not designed to analyze the interactions between such groups. Identifying these interactions gives an insight into the way different communities exchange messages or information, and allows the social network researcher to identify key actors, roles, and paths from one community to another. This article presents a novel model to use, in a conjoint way, the semantic information from the social network and its structural information to, first, find structurally and semantically related groups of nodes, and second, a layout algorithm for clustered graphs which divides the nodes into two types, one for nodes with edges connecting other communities and another with nodes connecting nodes only within their own community. With this division the visualization tool focuses on the connections between groups facilitating deep studies of augmented social networks.

Most of the layout algorithms for clustered graphs have been designed to differentiate the groups within the graph, however they do not take into account the interactions between such groups. Identifying these interactions allows to understand how the different communities exchange messages or information, and allows the social network researcher to identify key actors, social roles and paths from one community to another. These interactions are performed by actors who actually share links with different communities. Thus, dividing the set of nodes into those who interact with several communities and those who only interact with their own communities, will give some insight about how the communities interact.

In this paper, we show that transitive reduction can be used in real, large social networks analysis. Transitive reduction is an edge-removing operation on directed graphs that preserves some important properties and structures of the graph. After a presentation of this tool interest, we show its properties and how to use it. An application of the process on a real-word large social network (public and famous Enron corpus) built upon interaction data (email boxes) validates this approach.