Mahdi Jalili

Information cascades in complex networks

Information cascades are important dynamical processes in complex networks. An information cascade can describe the spreading dynamics of rumour, disease, memes, or marketing campaigns, which initially start from a node or a set of nodes in the network. If conditions are right, information cascades rapidly encompass large parts of the network, thus leading to epidemics or epidemic spreading. Certain network topologies are particularly conducive to epidemics, while others decelerate and even prohibit rapid information spreading. Here we review models that describe information cascades in complex networks, with an emphasis on the role and consequences of node centrality. In particular, we present simulation results on sample networks that reveal just how relevant the centrality of initiator nodes is on the latter development of an information cascade, and we define the spreading influence of a node as the fraction of nodes that is activated as a result of the initial activation of that node. A systemic review of existing results shows that some centrality measures, such as the degree and betweenness, are positively correlated with the spreading influence, while other centrality measures, such as eccentricity and the information index, have negative correlation. A positive correlation implies that choosing a node with the highest centrality value will activate the largest number of nodes, while a negative correlation implies that the node with the lowest centrality value will have the same effect.We discuss possible applications of these results, and we emphasize how information cascades can help us identify nodes with the highest spreading capability in complex networks.

Mesoscopic analysis of online social networks: the role of negative ties

A class of networks are those with both positive and negative links. In this manuscript, we studied the interplay between positive and negative ties on mesoscopic level of these networks, i.e., their community structure. A community is considered as a tightly interconnected group of actors; therefore, it does not borrow any assumption from balance theory and merely uses the well-known assumption in the community detection literature. We found that if one detects the communities based on only positive relations (by ignoring the negative ones), the majority of negative relations are already placed between the communities. In other words, negative ties do not have a major role in community formation of signed networks. Moreover, regarding the internal negative ties, we proved that most unbalanced communities are maximally balanced, and hence they cannot be partitioned into k nonempty sub-clusters with higher balancedness (k≥2). Furthermore, we showed that although the mediator triad ++- (hostile-mediator-hostile) is underrepresented, it constitutes a considerable portion of triadic relations among communities. Hence, mediator triads should not be ignored by community detection and clustering algorithms. As a result, if one uses a clustering algorithm that operates merely based on social balance, mesoscopic structure of signed networks significantly remains hidden.

Functional Brain Networks: Does the Choice of Dependency Estimator and Binarization Method Matter?

The human brain can be modelled as a complex networked structure with brain regions as individual nodes and their anatomical/functional links as edges. Functional brain networks are constructed by first extracting weighted connectivity matrices, and then binarizing them to minimize the noise level. Different methods have been used to estimate the dependency values between the nodes and to obtain a binary network from a weighted connectivity matrix. In this work we study topological properties of EEG-based functional networks in Alzheimer’s Disease (AD). To estimate the connectivity strength between two time series, we use Pearson correlation, coherence, phase order parameter and synchronization likelihood. In order to binarize the weighted connectivity matrices, we use Minimum Spanning Tree (MST), Minimum Connected Component (MCC), uniform threshold and density-preserving methods. We find that the detected AD-related abnormalities highly depend on the methods used for dependency estimation and binarization. Topological properties of networks constructed using coherence method and MCC binarization show more significant differences between AD and healthy subjects than the other methods. These results might explain contradictory results reported in the literature for network properties specific to AD symptoms. The analysis method should be seriously taken into account in the interpretation of network-based analysis of brain signals.

A trust-aware recommendation method based on Pareto dominance and confidence concepts

We proposed a trust-based collaborative filtering method called CPD.Implicit trust statements are identified and used in the recommendation process.A reliability measure is used to evaluate quality of implicit trust statements.Most prominent users are identified using Pareto dominance and confidence concepts.The results show that our method outperformed several state-of-the-art methods. Recommender systems are widely used to provide e-commerce users appropriate items. Collaborative filtering is one of the most successful recommender approaches which recommends items to a given user based on the opinions of his/her like-minded neighbors. However, the user-item ratings matrix, which is used as an input to the recommendation algorithm, is often highly sparse, leading to unreliable predictions. Recent studies demonstrated that information from social networks such as trust statements can be employed to improve accuracy of recommendations. However, there are not explicit trust relationships between most of users in many e-commerce applications. In this manuscript, we propose a method to identify implicit trust statements by applying a specific reliability measure. The Pareto dominance and confidence concepts are used to identify the most prominent users of which opinions are employed in the recommendation process. The proposed recommendation algorithm shows significant improvements in terms of accuracy and coverage measures as compared to the state-of-the-art recommenders.

Community detection in signed networks: The role of negative ties in different scales

Extracting community structure of complex network systems has many applications from engineering to biology and social sciences. There exist many algorithms to discover community structure of networks. However, it has been significantly under-explored for networks with positive and negative links as compared to unsigned ones. Trying to fill this gap, we measured the quality of partitions by introducing a Map Equation for signed networks. It is based on the assumption that negative relations weaken positive flow from a node towards a community, and thus, external (internal) negative ties increase the probability of staying inside (escaping from) a community. We further extended the Constant Potts Model, providing a map spectrum for signed networks. Accordingly, a partition is selected through balancing between abridgment and expatiation of a signed network. Most importantly, multi-scale spectrum of signed networks revealed how informative are negative ties in different scales, and quantified the topological placement of negative ties between dense positive ones. Moreover, an inconsistency was found in the signed Modularity: as the number of negative ties increases, the density of positive ties is neglected more. These results shed lights on the community structure of signed networks.

Link prediction in multiplex online social networks

Online social networks play a major role in modern societies, and they have shaped the way social relationships evolve. Link prediction in social networks has many potential applications such as recommending new items to users, friendship suggestion and discovering spurious connections. Many real social networks evolve the connections in multiple layers (e.g. multiple social networking platforms). In this article, we study the link prediction problem in multiplex networks. As an example, we consider a multiplex network of Twitter (as a microblogging service) and Foursquare (as a location-based social network). We consider social networks of the same users in these two platforms and develop a meta-path-based algorithm for predicting the links. The connectivity information of the two layers is used to predict the links in Foursquare network. Three classical classifiers (naive Bayes, support vector machines (SVM) and K-nearest neighbour) are used for the classification task. Although the networks are not highly correlated in the layers, our experiments show that including the cross-layer information significantly improves the prediction performance. The SVM classifier results in the best performance with an average accuracy of 89%.

Structural covariance of superficial white matter in mild Alzheimer's disease compared to normal aging

Interindividual variations in regional structural properties covary across the brain, thus forming networks that change as a result of aging and accompanying neurological conditions. The alterations of superficial white matter (SWM) in Alzheimers disease (AD) are of special interest, since they follow the AD‐specific pattern characterized by the strongest neurodegeneration of the medial temporal lobe and association cortices.

Finding the Most Influential Nodes in Pinning Controllability of Complex Networks

Identifying the best drivers (i.e., the nodes to apply the control signals in a large complex network), which gives the fastest synchronization to the reference state, is a challenge in pinning control of a network. There is not yet a method that exactly predicts a set of best drivers. In this brief, we introduce a novel method that gives first-order approximation for the importance of nodes in pinning control. A spectral measure (the largest eigenvalue of the augmented Laplacian matrix divided by the smallest eigenvalue) is considered as a pinning controllability metric. We develop this method based on the sensitivity analysis of the Laplacian eigenratio, resulting in the scoring of nodes based on their importance in pinning control. The method is rather simple to compute and needs a single eigendecomposition of the Laplacian matrix of the connection graph. Applying this technique on a number of model networks reveals its effectiveness over heuristic approaches.

Optimizing Dynamical Network Structure for Pinning Control.

Controlling dynamics of a network from any initial state to a final desired state has many applications in different disciplines from engineering to biology and social sciences. In this work, we optimize the network structure for pinning control. The problem is formulated as four optimization tasks: i) optimizing the locations of driver nodes, ii) optimizing the feedback gains, iii) optimizing simultaneously the locations of driver nodes and feedback gains, and iv) optimizing the connection weights. A newly developed population-based optimization technique (cat swarm optimization) is used as the optimization method. In order to verify the methods, we use both real-world networks, and model scale-free and small-world networks. Extensive simulation results show that the optimal placement of driver nodes significantly outperforms heuristic methods including placing drivers based on various centrality measures (degree, betweenness, closeness and clustering coefficient). The pinning controllability is further improved by optimizing the feedback gains. We also show that one can significantly improve the controllability by optimizing the connection weights.

TCARS: Time- and Community-Aware Recommendation System

Abstract With the abundance of information produced by users on items (e.g., purchase or rating histories), recommender systems are a major ingredient of online systems such as e-stores and service providers. Recommendation algorithms use information available from users–items interactions and their contextual data to provide a list of potential items for each user. These algorithms are constructed based on similarity between users and/or items (e.g., a user is likely to purchase the same items as his/her most similar users). In this work, we introduce a novel time-aware recommendation algorithm that is based on identifying overlapping community structure among users. Users’ interests might change over time, and accurate modeling of dynamic users’ preferences is a challenging issue in designing efficient personalized recommendation systems. The users–items interaction network is often highly sparse in real systems, for which many recommenders fail to provide accurate predictions. The proposed overlapping community structure amongst the users helps in minimizing the sparsity effects. We apply the proposed algorithm on two real-world benchmark datasets and show that it overcomes these challenges. The proposed algorithm shows better precision than a number of state-of-the-art recommendation methods.