Ranking-based Collaborative Clustering for Heterogeneous Information Network

Accurate user representation learning has been proven fundamental for many social media applications, including community detection, recommendation, etc. A major challenge lies in that, the available data in a single social network are usually very limited and sparse. In real life, many people are members of several social networks in the same time. Constrained by the features and design of each, any single social platform offers only a partial view of a user from a particular perspective. In this paper, we propose MV-URL, a multi-view user representation learning model to enhance user modeling by integrating the knowledge from various networks. Different from the traditional network embedding frameworks where either the whole framework is single-network based or each network involved is a homogeneous network, we focus on multiple social networks and each network in our task is a heterogeneous network. It's very challenging to effectively fuse knowledge in this setting as the fusion depends upon not only the varying relatedness of information sources, but also the target application tasks. MV-URL focuses on two tasks: user account linkage (i.e., to predict the missing true user account linkage across social media) and user attribute prediction. Extensive evaluations have been conducted on two real-world collections of linked social networks, and the experimental results show the superiority of MV-URL compared with existing state-of-art embedding methods. It can be learned online, and is trivially parallelizable. These qualities make it suitable for real world applications.

Most Americans are sorted into social networks that are largely politically homogeneous. A large body of political science research has explored the behavioral implications of being embedded in a politically homogeneous or heterogeneous network, but substantially less attention has been given to explaining why some people find themselves in politically homogeneous or heterogeneous social networks. In this article, we explore the psychological and physiological underpinnings of political network homogeneity. We use social network data from an original survey of 129 undergraduates paired with lab experimental evidence that measures individuals' physiological reactivity to an anticipated political discussion. Using our original survey and a separate nationally representative survey, we find suggestive evidence that individuals who are more socially anxious are more likely to share partisanship with their social network ties. Moreover, we find that individuals who experienced a greater increase in heart rate when anticipating a political discussion were more likely to be in homogeneous discussion networks, but we do not find a relationship between electrodermal activity and network homogeneity. Aversion to psychological and physiological discomfort induced by political discussions could contribute to social polarization in the American public.

The social network is a powerful data structure allowing the study of relationship information between entities. The existing methods only rely on analyzing homogeneous social network assuming only a single type of node and relation. Though, real world complex networks are not so. Usually it is a heterogeneous social network which assumes a network with different types of nodes and relation. A Multi Relational Network (MRN) is used to describe the nodes and links in the form of a directional semantic graph where each node is related with more than one relationship with others. The profile for each node can be generated based on the relation sequence of each node. To analyze the relation of node in the heterogeneous social network, role analysis is done in the profile creation. The evaluations are conducted on a real-world movie dataset with promising results.

Heterogeneous information networks have drawn much attention in recent years due to their significant applications, such as text mining, e-commerce, social networks, and bioinformatics. Clustering different types of objects simultaneously based upon not only their relations of the same type, but also the relations between different types of objects can improve the clustering quality mutually. In this paper, we propose a general model, in which both the homogeneous and heterogeneous relations are considered simultaneously, to describe the structure of the heterogeneous information networks and devise a novel parametric free multi-type overlapped clustering approach. In this model, different types of relations between different types of objects are represented by a group of matrices. In this way, we transfer the multi-type clustering problem into the information compression problem. Subsequently, greedy search approaches, which aim at describing the group of relational matrices with least bits, are proposed. Moreover, by discovering the discriminative clusters among different types of objects, we devise effective parameter-free strategies to discover either overlapping or non-overlapping structure among different types of clusters. Extensive experiments on real-world and synthetic data sets demonstrate our methods are effective and efficient.

The development of network technology has created various platforms and methods for information dissemination. When rumors spread in social networks, they will rapidly spread and may cause social harm. Also, there are groups in social networks that create and spread rumors for the purpose of profit, thus expanding the scope of rumors. Therefore, based on the theory of complex network propagation dynamics, the study of the propagation law of rumors and the design of effective prevention and control strategies is of practical importance and theoretical significance for understanding the propagation laws of rumors and controlling the outbreak of rumors. The spreading process of rumors on social network platforms is focused here. The intentional spreader based on the classic rumor‐spreading model is introduced. First, 2SIR rumor‐spreading models on homogeneous and heterogeneous networks are established, respectively. Second, the steady‐state analysis was separately carried out, and the corresponding propagation critical value was obtained: in the homogeneous network, the condition for the large‐scale spread of rumors is ; in the heterogeneous network, the condition for the large‐scale spread of rumors is . Finally, the simulation calculation and model feasibility verification were carried out on the model. The results show that the theoretical propagation threshold corresponds with the simulation results. According to the simulation results, the final influence of rumors has significantly decreased with decreasing values of β (intentional spreading rate) instead of α (unintentional spreading rate). It can be concluded that in the real‐life rumor control process, more resources need to be invested in reducing the rate of intentional transmission instead of being indiscriminately put on controlling all spreaders of rumors.

We study the dynamics of social networks in terms of population growth and control of user behavior. Most of the current research in social networks focus on static analysis through graph theoretic models to represent the networks or focus on modeling the traffic. Here, we study the cost of collaborative vs individualistic behavior of users in order to grow their network size in a social network. Each user incurs a cost (monetary or emotional) for collaboration. We formulate the behavior of the users as a non-linear optimization problem with a cost. The objective function of the optimization problem is obtained using a stochastic analysis of population growth in social networks, based on the first-passage time of a birth-death process. The stochastic model is validated by comparison with real data obtained from Twitter Results indicate that a homogeneous social network (in which users have similar characteristics) will be individualistic. However, heterogeneous social networks (users with different characteristics) exhibit a threshold effect, i.e., there is a minimum cost, below which the network is as collaborative as desired and a maximum cost above which the network is individualistic as required. To the best of our knowledge, this is one of the first analysis of dynamics of user behavior and temporal population growth in social networks.

The ability to generate innovations and capture the rents from innovation are important for firms’ competitive advantage. Increasingly firms seek knowledge abundant locations, or industry clusters, to access novel knowledge and generate innovations through knowledge recombinations (Schumpeter, 1934). We examine how different types of clusters impact on the innovation output, the knowledge flows among the clustered firms and, ultimately, on who captures the rents from innovation. The type of cluster reflects the configuration of firms and the interactions among firms, individuals and agencies in the cluster and is likely to be a major driver of both the innovative output and of which firms will be more likely to capture the rents from innovation. Extant research has noted that the social and business networks binding firms in clusters are excellent vehicles for the flow of knowledge that eases innovations, but different types of clusters may lead to different outcomes.

Harnessing heterogeneous social networks for better recommendations: A grey relational analysis approach

People often express opinions towards others in a social network, causing sentiment links to form among users. To develop effective methods for discovering implicit sentiment links among users, the extraction and exploitation of structural semantic information from heterogeneous social networks are of great importance. We propose a novel heterogeneous social network embedding-based approach for sentiment link prediction that takes both global structural information with multi-dimensional relations and heterogeneous context information into consideration to leverage rich and intrinsic association information. Specifically, the attributed heterogeneous social network and Sentic LSTM-based sentiment link network are employed to incorporate various explicit context knowledge and implicit multi-dimensional user interaction association knowledge into representation learning and sentiment link prediction. The experimental results on a real-world dataset show that the proposed approach has advantages over the state-of-the-art baselines. The results show the effectiveness of incorporating social relations and profile context information into sentiment link prediction, especially in cold-start scenarios. The learned embedding representation that incorporates both structural information with multi-dimensional relations and context information from heterogeneous social networks can improve sentiment link prediction performance. The proposed approach is effective and feasible for detecting unobserved sentiment links from online social networks and outperforms the state-of-the-art baselines in sentiment link prediction tasks.

On the learning of image social relevance from heterogeneous social network

Both homophily and heterophily are observed in humans. Homophily reinforces homogeneous social networks, and heterophily creates new experiences and collaborations. However, at the extremes, high levels of homophily can cultivate prejudice toward out-groups, whereas high levels of heterophily can weaken in-group support. Using data from 24,726 adults (M = 46 years; selected from 10,398 English neighborhoods) and the composition of their social networks based on age, ethnicity, income, and education, we tested the hypothesis that a middle ground between homophily and heterophily could be the most beneficial for individuals. We found that network homophily, mediated by perceived social cohesion, is associated with higher levels of subjective well-being but that there are diminishing returns, because at a certain point increasing network homophily is associated with lower social cohesion and, in turn, lower subjective well-being. Our results suggest that building diverse social networks provides benefits that cannot be attained by homogeneous networks.

Heterogeneous social networks, which are characterized by diverse interaction types, have resulted in new challenges for missing link prediction. Most deep learning models tend to capture type-specific features to maximize the prediction performances on specific link types. However, the types of missing links are uncertain in heterogeneous social networks; this restricts the prediction performances of existing deep learning models. To address this issue, we propose a multi-type transferable method ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MTTM</i> ) for missing link prediction in heterogeneous social networks, which exploits adversarial neural networks to remain robust against type differences. It comprises a generative predictor and a discriminative classifier. The generative predictor can extract link representations and predict whether the unobserved link is a missing link. To generalize well for different link types to improve the prediction performance, it attempts to deceive the discriminative classifier by learning transferable feature representations among link types. In order not to be deceived, the discriminative classifier attempts to accurately distinguish link types, which indirectly helps the generative predictor judge whether the learned feature representations are transferable among link types. Finally, the integrated <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MTTM</i> is constructed on this minimax two-player game between the generative predictor and discriminative classifier to predict missing links based on transferable feature representations among link types. Extensive experiments show that the proposed <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MTTM</i> can outperform state-of-the-art baselines for missing link prediction in heterogeneous social networks.

The online social network alignment problem aims at inferring the anchor links connecting the shared users across social networks, which are usually subject to the one-to-one cardinality constraint. Several existing social network alignment models have been proposed, many of which are based on the supervised learning setting. Given a set of labeled anchor links, a group of features can be extracted manually for the anchor links to build these models. Meanwhile, such methods may encounter great challenges in the application on real-world social network datasets, since manual feature extraction can be extremely expensive and tedious for the social networks involving heterogeneous information. In this paper, we propose to address the heterogeneous social network alignment problem with a deep learning model, namely DETA (Deep nETwork Alignment). Besides a small number of explicit features, DETA can automatically learn a set of latent features from the heterogeneous information. DETA models the anchor link one-to-one cardinality constraint as a mathematical constraint on the node degrees. Extensive experiments have been done on real-world aligned heterogeneous social network datasets, and the experimental results have demonstrated the effectiveness of the proposed model compared against the existing state-of-the-art baseline methods.

Unsupervised domain adaptation is mainly focused on the tasks of transferring knowledge from a fully-labeled source domain to an unlabeled target domain. However, in some scenarios, the labeled data are expensive to collect, which cause an insufficient label issue in the source domain. To tackle this issue, some works have focused on few-shot unsupervised domain adaptation (FUDA), which transfers predictive models to an unlabeled target domain through a source domain that only contains a few labeled samples. Yet the relationship between labeled and unlabeled source domains are not well exploited in generating pseudo-labels. Additionally, the few-shot setting further prevents the transfer tasks as an excessive domain gap is introduced between the source and target domains. To address these issues, we newly proposed an adaptive dispersal and collaborative clustering (ADCC) method for FUDA. Specifically, for the shortage of the labeled source data, a collaborative clustering algorithm is constructed that expands the labeled source data to obtain more distribution information. Furthermore, to alleviate the negative impact of domain-irrelevant information, we construct an adaptive dispersal strategy that introduces an intermediate domain and pushes both the source and target domains to this intermediate domain. Extensive experiments on the Office31, Office-Home, miniDomainNet, and VisDA-2017 datasets showcase the superior performance of ADCC compared to the state-of-the-art FUDA methods.

Most real-world information networks, such as social networks, are heterogeneous and as such, relationships in these networks can be of different types and hence carry differing semantics. Therefore techniques for link prediction in homogeneous networks cannot be directly applied on heterogeneous ones. On the other hand, works that investigate link prediction in heterogeneous networks do not necessarily consider network dynamism in sequential time intervals. In this work we propose a technique that leverages a combination of latent and topological features to predict a target relationship between two nodes in a dynamic heterogeneous information network. Our technique, called MetaDynaMix, effectively combines meta path-based topology features and inferred latent features that incorporate temporal network changes in order to capture network (1) heterogeneity and (2) temporal evolution, when making link predictions. Our experiment results on two real-world datasets show statistically significant improvement over AUCROC and prediction accuracy compared to the state of the art techniques.