Table 2 - uploaded by David Sivakoff
Content may be subject to copyright.
Source publication
A number of real world problems in many domains (e.g. sociology, biology, political science and communication networks) can be modeled as dynamic networks with nodes representing entities of interest and edges representing interactions among the entities at different points in time. A common representation for such models is the snapshot model - wh...
Context in source publication
Context 1
... (Fig b-top) and EM-KL (Fig b-bottom) have the smallest variance, but DeltaCon has two false negatives at 4 and 5 . Figure 4 Order Window Index Type of Change 1 15 The weight sequence of 1/3 of the nodes is re-generated 2 30 The weight sequence of 2/3 of the nodes is re-generated 3 60 Half of the communities change their (inter-and intra-community) connection rate, overall density retained 4 75 All of the communities change their (inter-and intra-community) connection rate, overall density retained 5 90 ...
Similar publications
Christopher Alexander published his longest and arguably most philosophical work, The Nature of Order, beginning in 2003. Early criticism assessed that text to be a speculative failure; at best, unrelated to Alexander’s earlier, mathematically grounded work. On the contrary, this review presents evidence that the newer work was a logically consiste...
Citations
... The results illustrate that when experiencing clustering events, there is a transition in the time scale (from slow to fast) and direction (from hierarchical to distributed) of information transfer in the network. Wang et al. [48] expressed the evolution of the temporal network as a Markov network and detected change points through estimating and comparing the joint edge (dyad) distribution. Experiments on the Senate cosponsorship network show that the method is more efficient than the other approaches in the same period while ensuring a good detection effect. ...
The social network is closely related to people’s lives. And social events are the products of the human subjective initiative during the evolution of networks. Therefore, there is a close correlation between social events and network evolution. This paper studies the characteristics of network evolution corresponding to social events from the perspective of temporal networks. The change point detection method is applied to capture the “shocks” of social events on the network structure. Then, the patterns of structural changes are analyzed based on the theory of community evolution. Experiments on two cases illustrate that social events are significant milestones to promote the development of social networks. And the mesostructure is the intermediary connecting evolving network and social events.
... Past research for identifying change points used stochastic models, of either scalar values representing the longitudinal data [5], or probabilistic and model-based representations of the network [3,[6][7][8]. However, the works mentioned above did not examine the complex network's structure as manifested through distributions. ...
... A complementary approach, similar to ours, is to extract a large number of features from each consecutive graph snapshots and find the distance between them [6,7,16,24]. A change is determined if a predefined threshold for the distance is crossed. ...
... Frameworks for change point detection divide the data to consecutive snapshots according to a natural division derived from the nature of the data, such as daily or weekly snapshots of organizational frameworks, or monthly graphs of votes. In methods measuring the distance between features extracted from two consecutive graph snapshots [6,7,16], a change is Hypothesis testing over a distance measure is used to determine whether the underlying model has changed. On the left graphs generated from the same model, on the right a graph generated from a changed model. ...
Changes in the structure of observed social and complex networks can indicate a significant underlying change in an organization, or reflect the response of the network to an external event. Automatic detection of change points in evolving networks is rudimentary to the research and the understanding of the effect of such events on networks. Here we present an easy-to-implement and fast framework for change point detection in evolving temporal networks. Our method is size agnostic, and does not require either prior knowledge about the network’s size and structure, nor does it require obtaining historical information or nodal identities over time. We tested it over both synthetic data derived from dynamic models and two real datasets: Enron email exchange and AskUbuntu forum. Our framework succeeds with both precision and recall and outperforms previous solutions.
... The aim is to propose flexible methods that can adapt to data that evolve or change over time. A similar problem found in the complex network literature is the change point detection [22][23][24][25] , which seeks to detect the moments of change between one concept (or network pattern) to the other. In general, this approach focus on measuring the dissimilarity between one snapshot of the temporal network and the next one in time. ...
... In general, this approach focus on measuring the dissimilarity between one snapshot of the temporal network and the next one in time. A change is detected when the dissimilarity passes a fixed threshold 22,23 . These models usually depend on a set of time frames, for which a solution is proposed in 24 . ...
Temporal network mining tasks are usually hard problems. This is because we need to face not only a large amount of data but also its non-stationary nature. In this paper, we propose a method for temporal network pattern representation and pattern change detection following the reductionist approach. The main idea is to model each stable (durable) state of a given temporal network as a community in a sampled static network and the temporal state change is represented by the transition from one community to another. For this purpose, a reduced static single-layer network, called a target network, is constructed by sampling and rearranging the original temporal network. Our approach provides a general way not only for temporal networks but also for data stream mining in topological space. Simulation results on artificial and real temporal networks show that the proposed method can group different temporal states into different communities with a very reduced amount of sampled nodes.
... In [15], the authors proposed the first change-point detection method for evolving networks that uses generative network models and statistical hypothesis testing. Wang et al. [19] proposed a hierarchical change point detection method to detect both inter-community(local change) and intra-community(global change) evolution. A recent work by Masuda et al. [11] used graph distance measures and hierarchical clustering to identify sequences of system state dynamics. ...
Link streams model interactions over time in a wide range of fields. Under this model, the challenge is to mine efficiently both temporal and topological structures. Community detection and change point detection are one of the most powerful tools to analyze such evolving interactions. In this paper, we build on both to detect stable community structures by identifying change points within meaningful communities. Unlike existing dynamic community detection algorithms, the proposed method is able to discover stable communities efficiently at multiple temporal scales. We test the effectiveness of our method on synthetic networks, and on high-resolution time-varying networks of contacts drawn from real social networks.
... Polynomial-time consistent estimation for unseeded graphon estimation enables rich future inferences, such as network comparison (Asta and Shalizi, 2014;Lei, 2018b), independence test and correlation estimation (Lyzinski, 2018), change point detection (Wang et al., 2017), joint estimations of common structural parameters by combining information from several networks (Le and Levina, 2017;Liu et al., 2018) and so on. It is also of both theoretical and practical interest to analyze the asymptotic limiting distribution for full-rank graphons, analogous to existing results under low-rank models including Tang and Priebe (2016) and Levin et al. (2017). ...
We propose a consistent polynomial-time method for the unseeded node matching problem for networks with smooth underlying structures. Despite widely conjectured by the research community that the structured graph matching problem to be significantly easier than its worst case counterpart, well-known to be NP-hard, the statistical version of the problem has stood a challenge that resisted any solution both provable and polynomial-time. The closest existing work requires quasi-polynomial time. Our method is based on the latest advances in graphon estimation techniques and analysis on the concentration of empirical Wasserstein distances. Its core is a simple yet unconventional sampling-and-matching scheme that reduces the problem from unseeded to seeded. Our method allows flexible efficiencies, is convenient to analyze and potentially can be extended to more general settings. Our work enables a rich variety of subsequent estimations and inferences.
... Change-point and anomaly detection for temporal networks is also concerned with detecting changes of networks over time 7,[17][18][19][20][21] . A main difference between our system state dynamics and these methods is that detection of system state dynamics is concerned with not only the change, but what is before and after the change. ...
Many time-evolving systems in nature, society and technology leave traces of the interactions within them. These interactions form temporal networks that reflect the states of the systems. In this work, we pursue a coarse-grained description of these systems by proposing a method to assign discrete states to the systems and inferring the sequence of such states from the data. Such states could, for example, correspond to a mental state (as inferred from neuroimaging data) or the operational state of an organization (as inferred by interpersonal communication). Our method combines a graph distance measure and hierarchical clustering. Using several empirical data sets of social temporal networks, we show that our method is capable of inferring the system's states such as distinct activities in a school and a weekday state as opposed to a weekend state. We expect the methods to be equally useful in other settings such as temporally varying protein interactions, ecological interspecific interactions, functional connectivity in the brain and adaptive social networks.
This survey paper presents a comprehensive and conceptual overview of anomaly detection using dynamic graphs. We focus on existing graph-based anomaly detection (AD) techniques and their applications to dynamic networks. The contributions of this survey paper include the following: i) a comparative study of existing surveys on anomaly detection; ii) a D ynamic G raph-based A nomaly D etection ( DGAD ) review framework in which approaches for detecting anomalies in dynamic graphs are grouped based on traditional machine-learning models, matrix transformations, probabilistic approaches, and deep-learning approaches; iii) a discussion of graphically representing both discrete and dynamic networks; and iv) a discussion of the advantages of graph-based techniques for capturing the relational structure and complex interactions in dynamic graph data. Finally, this work identifies the potential challenges and future directions for detecting anomalies in dynamic networks. This DGAD survey approach aims to provide a valuable resource for researchers and practitioners by summarizing the strengths and limitations of each approach, highlighting current research trends, and identifying open challenges. In doing so, it can guide future research efforts and promote advancements in anomaly detection in dynamic graphs.
How can we detect traffic disturbances from international flight transportation logs, or changes to collaboration dynamics in academic networks? These problems can be formulated as detecting anomalous change points in a dynamic graph. Current solutions do not scale well to large real world graphs, lack robustness to large amount of node additions / deletions and overlook changes in node attributes. To address these limitations, we propose a novel spectral method: Scalable Change Point Detection (SCPD). SCPD generates an embedding for each graph snapshot by efficiently approximating the distribution of the Laplacian spectrum at each step. SCPD can also capture shifts in node attributes by tracking correlations between attributes and eigenvectors. Through extensive experiments using synthetic and real world data, we show that SCPD (a) achieves state-of-the-art performance, (b) is significantly faster than the state-of-the-art methods and can easily process millions of edges in a few CPU minutes, (c) can effectively tackle a large quantity of node attributes, additions or deletions and (d) discovers interesting events in large real world graphs. Code is publicly available at https://github.com/shenyangHuang/SCPD.git.KeywordsAnomaly DetectionDynamic GraphsSpectral Methods
