Global Network Information Research Articles

Graph regularized nonnegative matrix factorization for link prediction in directed temporal networks using PageRank centrality

Temporal link prediction aims to predict new links at time T + 1 based on a given temporal network from time 1 to T. Some existing temporal link prediction methods only consider the naturally links direction formation but ignore the local and global information of temporal networks. To address this issue, in this paper, we propose a novel graph regularized nonnegative matrix factorization algorithm for the problem of temporal link prediction by using PageRank centrality and asymmetric link clustering coefficient, referred to as the GNMFCA. Specifically, we use graph regularization to capture the local information of each slice in temporal networks and PageRank centrality to compute the importance of nodes in each slice, which captures the global information of each slice in temporal networks. By jointly optimizing them in a nonnegative matrix factorization model, the proposed model can preserve both the local and global information simultaneously. Besides, we propose effective multiplicative updating rules to solve the proposed model and provide the convergence analysis of the iterative algorithm. Finally, we perform numerical experiments on eight artificial and four real-world temporal networks to demonstrate that GNMFCA outperforms some existing temporal link prediction algorithms.

Identification of novel prognostic targets in glioblastoma using bioinformatics analysis

BackgroundGlioblastoma (GBM) is the most malignant grade of glioma. Highly aggressive characteristics of GBM and poor prognosis cause GBM-related deaths. The potential prognostic biomarkers remain to be demonstrated. This research builds up predictive gene targets of expression alterations in GBM utilizing bioinformatics analysis.Methods and resultsThe microarray datasets (GSE15824 and GSE16011) associated with GBM were obtained from Gene Expression Omnibus (GEO) database to identify the differentially expressed genes (DEGs) between GBM and non-tumor tissues. In total, 719 DEGs were obtained and subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) for function enrichment analysis. Furthermore, we constructed protein–protein Interaction (PPI) network among DEGs utilizing Search Tool for the Retrieval of Interacting Genes (STRING) online tool and Cytoscape software. The DEGs of degree > 10 was selected as hub genes, including 73 upregulated genes and 21 downregulated genes. Moreover, MCODE application in Cytoscape software was employed to identify three key modules involved in GBM development and prognosis. Additionally, we used the Gene expression profiling and interactive analyses (GEPIA) online tool to further confirm four genes involving in poor prognosis of GBM patients, including interferon-gamma-inducible protein 30 (IFI30), major histocompatibility complex class II-DM alpha (HLA-DMA), Prolyl 4-hydroxylase beta polypeptide (P4HB) and reticulocalbin-1 (RCN1). Furthermore, the correlation analysis indicated that the expression of IFI30, an acknowledged biomarker in glioma, was positively correlated with HLA-DMA, P4HB and RCN1. RCN1 expression was positively correlated with P4HB and HLA-DMA. Moreover, qRT-PCR and immunohistochemistry analysis further validated the upregulation of four prognostic markers in GBM tissues.ConclusionsAnalysis of multiple datasets combined with global network information and experimental verification presents a successful approach to uncover the risk hub genes and prognostic markers of GBM. Our study identified four risk- and prognostic-related gene signatures, including IFI30, HLA-DMA, P4HB and RCN1. This gene sets contribute a new perspective to improve the diagnostic, prognostic, and therapeutic outcomes of GBM.

Social media use and protection of children’s rights

Similar to traditional media, such as television and radio, new social media are becoming more influential and attractive every day. The goal of the present study was to expand the scope of previous reviews on child protection issues to focus on primary and secondary school students, ages when research indicates that children begin to use mobile phones and social media. The actual issue of legal regulation of the Kazakhstan Internet space raises in the article. The author examines the issues of combining the methods of legal regulation and self-regulation on the Internet. In particular, we are talking about the fact that the threat from the Internet for children is growing every day. Prospects of soft law in the international regulation of the Internet, in countries such as the United States, Great Britain and Russia are assessed from different points of view. Especially the author draws attention to such topics as cyberbullying, hate speech, pornography and encroachment on the honor and dignity of citizens. This can be considered as the main results of the researches.       The relevance of this topic is expressed in the fact that, the Internet is still not sufficiently regulated from the legal point of view. The constant increase in the number of subscribers and the growing importance of information exchange through the network attract the public's attention to the problems of regulation, the development of rules for the fair, legitimate functioning of the Internet by the state.       The author uses methods of collecting statistical data and necessary facts, their analysis, ranking, comparison, as well as infographic methods.      The main problems of legal regulation are related to the order and conditions of the use of telecommunications networks and protection of the rights and legitimate interests of various entities during the transfer of information in global computer networks. There is a need to develop an adequate national legislation with all modern requirements.      At present, the information legislation of Kazakhstan and other neighboring countries are very extensive. However, it is still far from perfect and requires working out and adoption of new acts. This is the value of this study.   Keywords: Internet, social media, legal regulation, children's rights, blocking of information, access to information, security.

Network Representation Learning Algorithm Based on Complete Subgraph Folding

Network representation learning is a machine learning method that maps network topology and node information into low-dimensional vector space. Network representation learning enables the reduction of temporal and spatial complexity in the downstream data mining of networks, such as node classification and graph clustering. Existing algorithms commonly ignore the global topological information of the network in network representation learning, leading to information loss. The complete subgraph in the network commonly has a community structure, or it is the component module of the community structure. We believe that the structure of the community serves as the revealed structure in the topology of the network and preserves global information. In this paper, we propose SF-NRL, a network representation learning algorithm based on complete subgraph folding. The algorithm preserves the global topological information of the original network completely, by finding complete subgraphs in the original network and folding them into the super nodes. We employ the network representation learning algorithm to study the node embeddings on the folded network, and then merge the embeddings of the folded network with those of the original network to obtain the final node embeddings. Experiments performed on four real-world networks prove the effectiveness of the SF-NRL algorithm. The proposed algorithm outperforms the baselines in evaluation metrics on community detection and multi-label classification tasks. The proposed algorithm can effectively generalize the global information of the network and provides excellent classification performance.

SDUNet: Road extraction via spatial enhanced and densely connected UNet

Extracting road maps from high-resolution optical remote sensing images has received much attention recently, especially with the rapid development of deep learning methods. However, most of these CNN based approaches simply focused on multi-scale encoder architectures or multiple branches in neural networks, and ignored some inherent characteristics of the road surface. In this paper, we design a novel network for road extraction based on spatial enhanced and densely connected UNet, called SDUNet. SDUNet aggregates both the multi-level features and global prior information of road networks by combining the strengths of spatial CNN-based segmentation and densely connected blocks. To enhance the feature learning about prior information of road surface, a structure preserving model is designed to explore the continuous clues in the spatial level. Experimental results on two benchmark datasets show that the proposed method achieves the state-of-the-art performance, compared with previous approaches for road extraction. Code will be made available on https://github.com/MrStrangerYang/SDUNet.

Multi-agent deep reinforcement learning for resilience-driven routing and scheduling of mobile energy storage systems

Extreme events are featured by high impact and low probability, which can cause severe damage to power systems. There has been much research focused on resilience-driven operational problems incorporating mobile energy storage systems (MESSs) routing and scheduling due to its mobility and flexibility. However, existing literature focuses on model-based optimization approaches to implement the routing process of MESSs, which can be time consuming and raise privacy issues since the requirement for global information of both power and transportation networks. Furthermore, a real-time automatic control scheme of MESSs has become a challenging task due to the system high variability. As such, this paper develops a model-free real-time multi-agent deep reinforcement learning approach featuring parameterized double deep Q-networks to reformulate the coordination effect of MESSs routing and scheduling process as a Partially Observable Markov Game, which is capable of capturing a hybrid policy including both discrete and continuous actions. A coupled transportation network and linearized AC-OPF algorithm are realized as the environment, while the internal uncertainties associated with renewable energy sources, load profiles, line outages, and traffic volumes are incorporated into the proposed data-driven approach through learning procedure. Extensive case studies including both 6-bus and 33-bus power networks are developed to evaluate the effectiveness of the proposed approach. Specifically, a detailed comparison between different multi-agent reinforcement learning and model-based optimization approaches is conducted to present the superior performance of the proposed approach.

Distributed control for state‐of‐charge balance and load voltage regulation in DC microgrids with clustered generations

AbstractIn this paper, the issue of state‐of‐charge balance and load voltage regulation of DC microgrids used in solar‐powered vehicles is investigated. The system consists of multiple distributed generation units, each of which is equipped with photovoltaic cells and energy storage systems. Based on the information shared among the neighbors, a single‐layered distributed control strategy is proposed to solve two aspects of the problem simultaneously. Rigorous analysis of the closed‐loop system ensures stability and the effectiveness of the control strategy. The power constraints of each distributed generation unit are taken into consideration so that safety and applicability are guaranteed. To avoid the cycling of the batteries, the state‐of‐charge balance is achieved by regulating the output power feeding a common load, promoting efficiency and battery life. The proposed controller operates without any knowledge about load, line impedance, and global information of communication network, and consequently possesses the capability of plug‐and‐play. Finally, the effectiveness and applicability are illustrated by an example from the Sky‐Sailor project.

Raze policy conflicts in SDN

Software Defined Networking (SDN) enables flexible network management with a well-defined abstraction between control and data plane. In this way, operators could issue the policies, e.g., forwarding path, flow counting and rate limiting, from the control plane, which will be enforced by the flow table rules in the data plane. However, multiple active policies with the same priority will potentially trigger conflicts among policies with overlapped flow space, causing the flow table explosion. In contrast to the local switch conflict resolution schemes proposed by previous works, this paper tackles the same problem from a different angle and resolves the policy conflict problem by coordinating all switches under a global centralized view. Specifically, we propose COnflict RAzor (CORA), which tremendously reduces the storage cost of conflicting policies leveraging the global network information obtained in the controller. The basic idea of CORA is migrating policies causing large explosions across the network if necessary, while keeping the semantics equivalence. We prove CORA’s NP hardness and propose a heuristic to efficiently search a near-optimal policy migration strategy. Our experiments demonstrate that, CORA can effectively reduce the flow table storage occupation by averagely 79.8% within less than 40 s, which is 47.9% more efficient than the state-of-the-art.

Fully distributed event-triggered time-varying formation control of multi-agent systems subject to mode-switching denial-of-service attacks

In this study, the time-varying state formation control problem for general linear multi-agent systems (MASs) subject to mode-switching denial-of-service (MSDoS) attacks is considered. Based on only the sampled state information of itself and neighboring agents at event-triggered instants, a novel fully distributed event-triggered secure control strategy without continuous communication between agents is delicately designed. Note that the control strategy is implemented in a fully distributed manner, which means that it does not require any global network information. By using the multiple Lyapunov function approach based on edge-dependent average dwell time, this study presents the MASs subject to MSDoS attacks with limited attack frequency and attack width can achieve the specified time-varying state formation structure under the designed control strategy. Furthermore, this study presents that in any finite time, the control strategy does not exhibit Zeno behavior. Finally, the effectiveness and performance of the designed control strategy are validated by a numerical simulation.

Event‐triggered adaptive sliding mode control for consensus of heterogeneous nonlinear multiagent systems

AbstractIn this article, we peruse the consensus problem of nonlinear heterogeneous multiagent systems (MASs) in the presence of uncertainty and external disturbances. A new event‐triggered adaptive sliding mode control is first proposed to achieve consensus for MASs consisting of heterogeneous nonlinear agents with nonidentical inherent dynamics and different‐order dynamics. It is supposed that the uncertainties and their derivatives are bounded, but the upper‐bounds are unknown. For reaching consensus in heterogeneous MAS, the proposed controller integrates two fundamental actions concurrently. First, an adaptive sliding‐mode control strategy is adopted to guarantee the robust finite‐time regulation of each individual nonlinear, uncertain, and heterogeneous agent dynamics to the sliding surface and keeping on it despite the unknown model uncertainties and disturbances. Then, the decoupled heterogeneous sliding surfaces are designed to align the agent trajectories to a consensus through local interactions, when they reach the sliding surface. Also, by applying a novel event‐based strategy, we facilitate the usage of limited resources of embedded microprocessors that actuate the controller updates. We propose centralized and distributed event‐triggered schemes, in which each agent decides itself when to transmit its current state to its neighbors based on the global network information or only the knowledge of their own neighbors' states. Moreover, it is shown that the interevent intervals are lower‐bounded by a positive constant, and Zeno behavior is excluded. In the end, by doing simulation, verification of the theoretical results are presented.

ECLP: Friend Recommendation Using Ensemble Approach for Detecting Communities Performing Link Prediction

Social networks provide a variety of online services that play an important role in new connections among members to share their favorite media, document, and opinions. For each member, these networks should precisely recommend (predict) the link of members with the highest common interests. Because of the huge volume of users with different types of information, these networks encounter challenges such as dispersion and accuracy of link prediction. Moreover, networks with numerous users have the problem of computational and time complexity. These problems are caused because all the network nodes contribute to calculations of link prediction and friend suggestions. In order to overcome these drawbacks, this paper presents a new link prediction scheme containing three phases to combine local and global network information. In the proposed manner, dense communities with overlap are first detected based on the ensemble node perception method which leads to more relevant nodes and contributes to the link prediction and speeds up the algorithm. Then, these communities are optimized by applying the binary particle swarm optimization method for merging the close clusters. It maximizes the average clustering coefficient (ACC) of the whole network which results in an accurate and precise prediction. In the last phase, relative links are predicted by Adamic/Adar similarity index for each node. The proposed method is applied to Astro-ph, Blogs, CiteSeer, Cora, and WebKB datasets, and its performance is compared to state-of-the-art schemes in terms of several criteria. The results imply that the proposed scheme has a significant accuracy improvement on these datasets.

Joint Throughput-Power Optimization of Fog-RAN Using Rate-Splitting Multiple Access and Reinforcement-Learning Based User Clustering

Rate-splitting multiple access (RSMA) and reinforcement-learning (RL) based user clustering are leveraged to manage interference of a downlink fog radio access network. In particular, clustering of the user devices (UDs) is considered. The UDs of a given cluster can receive data from a certain fog access point (F-AP) over the same radio resource blocks (RRBs) simultaneously. To this end, RSMA enabled data transmission is exploited at each UD cluster. To manage intra-cluster and inter-cluster interference of the network, we focus on the resource allocation to maximize the sum-rate and to minimize total transmission power of the F-APs in each transmission slot. Towards this goal, we optimize jointly the F-APs’ transmit power allocation for RSMA enabled data transmission, clustering of UDs, and assignment of RRBs among the F-APs. The proposed optimization problem is NP-hard, and as a result, a global optimal solution is computationally intractable even with the centralized implementation. To obtain an efficient solution without having the global network information, the proposed optimization problem is decomposed into two sub-problems, namely, UD clustering and resource allocation among the F-APs. Specifically, the UD clusters are obtained by applying a multi-agent RL technique, and the resource allocation among the F-APs is obtained by applying the fractional programming, Lagrangian duality, and alternating optimization techniques. A distributed user clustering-power allocation-RRB assignment (UC-PA-RA) algorithm is proposed, and its convergence to the near-optimal solution is proved. Through extensive simulations, the superiority of the proposed UC-PA-RA algorithm over the contemporary multiple access schemes, UD clustering technique, and RL method is verified.

Semiglobal Observer-Based Positive Scaled Edge-Consensus of Networked Discrete-Time Systems Under Actuator Saturation

In this article, by using the linear graph theory, modified algebraic Riccati equation (MARE)-based method and duality principle, two interaction protocols are proposed to achieve the scaled edge-consensus of networked discrete-time systems with unmeasurable edge-sates. Meanwhile, drawing support from MARE-based low-gain feedback technique and positive system theory, sufficient conditions are derived to ensure the bounded control inputs and positive edge-states. Moreover, feedback matrices and observer matrices are, respectively, constructed by utilizing the solutions of MARE and modified algebraic dual Riccati equation (MADRE). Remarkably, the theoretical results only use the edge number and vertex number of network rather than the global information of network. Finally, five examples are shown to demonstrate the feasibility of the interaction protocols.

Heterogeneous academic network embedding based multivariate random-walk model for predicting scientific impact

The prediction of current scientific impact of papers and authors has been extensively studied to help researchers find valuable papers and recent research directions, also help policymakers make recruitment decisions or funding allocation. However, how to accurately evaluate the future impact of them, especially for new papers and young researchers, is the focus of scientific impact prediction research, and is less explored. Existing graph-based methods heavily depend on the global structure information of heterogeneous academic network and ignore the local structure information and text information, which may provide important clues to identify influential papers and authors with novel perspective. In this paper, we propose a hybrid model called ESMR to predict the future influence of papers and authors by mainly exploiting these information mentioned above. Specifically, we first put forward a novel network embedding-based model, which can capture not only the local structure information, but also the text information of papers into a unified embedding representation. Then, the future impact of papers and authors is mutually ranked by integrating the learned embedding representations into a multivariate random-walk model. Empirical results on two real datasets demonstrate that the proposed method significantly outperforms the existing state-of-the-art ranking methods.

Consensus control for nonlinear multi-agent systems with event-triggered communications

This paper investigates the consensus problems for a class of nonlinear multi-agent systems (MASs) by distributed event-based control protocols. Firstly, we design the static distributed event-triggered control protocols to achieve leader-follower and leaderless consensus problems, respectively, and the systems exclude Zeno behavior under the proposed control schemes. In order to avoid utilizing the global information of the system networks, that is, the minimum eigenvalues of the correlation matrices. Then, the fully distributed event-based control schemes are designed by introducing the time-varying coupling weights to achieve the leader-follower and leaderless consensus problems, respectively. It has been shown that the Zeno behavior is ruled out in the MASs under the fully distributed event-based control protocols. All static and dynamic event-based control schemes are able to be constructed based on the solvability of linear matrix inequalities (LMIs). Finally, the rationality of the theoretical results is illustrated by the example.

Dynamic Triggering Mechanisms for Distributed Adaptive Synchronization Control and Its Application to Circuit Systems

Nonlinear couplings among units (nodes) are ubiquitous in engineering systems including, e.g., radar and sonar systems, which have been ignored in most works. In this article, the problem of distributed synchronization of nonlinear networked systems with nonlinear couplings is studied. Specifically, two kinds of nodes’ communication couplings including nonlinear relative and nonlinear absolute state couplings are considered. To reduce the requirements of control and communication among nodes and avoid any global network information, two edge-based fully adaptive event-triggered control protocols based on nonlinear relative and absolute state couplings are proposed by using the projection operator technique, which is followed by design of corresponding dynamic event-triggered mechanisms. The advantages of our proposed dynamic event-triggered strategies show that it can boil down to existing static ones as special examples, and the minimal inter-execution time of the proposed dynamic triggering laws is larger than that of static ones. Theoretical analysis shows that the proposed algorithm not only guarantees fully adaptive Zeno-free synchronization of networked systems without requiring any global information, but also avoids continuous communications among nodes, and considerably reduce the frequency of controller updates. Finally, the practical merits of the proposed algorithms are corroborated using a Chua’s circuit network.

Semi-global leader-following consensus-based formation flight of unmanned aerial vehicles

In this paper, we investigate a formation control problem of multi-agent systems (specifically a group of unmanned aerial vehicles) based on a semi-global leader-following consensus approach with both the leader and the followers subject to input saturation. Utilizing the low gain feedback design technique, a distributed static control protocol and a distributed adaptive control protocol are constructed. The former solves the problem under an assumption that the communication network is undirected, and it depends on the global information of the graph. For the latter, we relax the undirected graph to directed graph. Moreover, an adaptive updating gain is designed to avoid using the global information of the communication network. It is shown that the consensus protocols can solve the semi-global leader-following consensus problem if the leader agent is globally reachable. The results are verified successfully by both simulation and real flight tests.

Design of Distributed Event-Triggered Average Tracking Algorithms for Homogeneous and Heterogeneous Multiagent Systems

This article addresses the design problem of distributed event-triggered average tracking (DETAT) algorithms for homogeneous and heterogeneous multiagent systems. The objective of the DETAT problem is to develop a group of distributed cooperative control algorithms with event-triggered strategies for agents to track the average of multiple time-varying reference signals. First, for homogeneous linear multiagent systems, based on sampling measurements and model-relied holding techniques, a class of static-gain DETAT algorithms is proposed with a couple of local event-triggered functions for estimators and controllers, respectively. Compared with the existing distributed average tracking (DAT) algorithms, the static-gain DETAT algorithms greatly reduce the cost over communication networks and the frequency of control protocol updates. Second, to reduce the chattering phenomenon caused by nonsmooth items in static-gain algorithms and requirements of the global information of networks, smooth dynamic-gain DETAT algorithms are introduced based on boundary layer approximation methods and self-adaptive principles. Third, for heterogeneous linear multiagent systems, a new algorithm is established by using the output regulation techniques for the heterogeneous DETAT problem. The outputs of heterogeneous agents can ultimately track the average of multiple time-varying reference signals. To the best of our knowledge, it is the first time to study the DETAT problem for heterogeneous multiagent systems. Finally, some examples are presented to show the validity of theoretical results.

Node classification using kernel propagation in graph neural networks

In this work, we introduce a kernel propagation method that enables graph neural networks (GNNs) to leverage higher-order network structural information without increasing the complexity of the networks. Recent studies have introduced GNNs that include higher-order neighborhood features containing global network information by propagating node features using a higher-order feature propagation rule. Though these GNNs have shown to improve node classification performance, they fail to include local connectivity information. Alternatively, GNNs also concatenate increasing orders of adjacency matrix in deeper layers in order to include higher-order structural information. In addition to global network information, GNNs also make use of node features which are network and node dependent features that serve to distinguish structurally isomorphic sub-structures within graphs. However, such node features may not always be available or depending on the network, may lead to deteriorating classification performance. Hence, to resolve these limitations, we propose a kernel propagation method that introduces a pre-processing step for GNNs to leverage higher-order structural features. The higher-order structural features are computed using a weighted random walk matrix which is node independent while using the first-order spectral propagation rule which explicitly considers local connectivity. Through our benchmark experiments, we find that the computed higher-order structural features are capable of replacing node dependent features while performing node classification task with performance on par with the state of the art approaches. Further, we also find that including both node features and higher-order structural features increases the performance of GNNs on large scale benchmark networks considered in this study. Our results show that considering local and global structural information as input to GNNs lead to an improvement in node classification performance in the absence/presence of node features without loss of performance.

Identifying influential nodes in complex networks: Effective distance gravity model

The identification of important nodes in complex networks is an area of exciting growth due to its applications across various disciplines like disease control, data mining and network system control. Many measures have been proposed to date, but they are either based on the locality of nodes or the global nature of the network. These measures typically use the traditional Euclidean Distance, which only focuses on local static geographic distance between nodes but ignores the dynamic interaction between nodes in real-world networks. Both the static and dynamic information should be considered for the purpose of identifying influential nodes. In order to address this problem, we have proposed an original and novel gravity model with effective distance for identifying influential nodes based on information fusion and multi-level processing. Our method is able to comprehensively consider the global and local information of complex networks, and also utilizes the effective distance to incorporate static and dynamic information. Moreover, the proposed method can help us mine for hidden topological structure of real-world networks for more accurate results. The susceptible infected model, Kendall correlation coefficient and eight existing identification methods are utilized to carry out simulations on twelve different real networks.