Weighted Personalized Factorizations for Network Classification with Approximated Relation Weights

In this paper, we first put forward a mathematical model to find a path between two nodes in network based on Brownian particles’ random walk from multiple sources. The performance analysis proves that this model is as good as traditional multiple Brownian particles’ random walk. Then we propose a method to find the efficient path between two nodes in wireless sensor networks according to this mathematical model. The core of the method is to divide the nodes in WSN into two different types: general type and base-station type. The general type nodes forwards messages randomly to one of its’ neighbors and the base-station type nodes broadcast messages to all of its’ neighbors. The experiment shows that after huge number of simulations this way could find the efficient path between two nodes with high probability.

BackgroundTraditional drug research and development is high cost, time-consuming and risky. Computationally identifying new indications for existing drugs, referred as drug repositioning, greatly reduces the cost and attracts ever-increasing research interests. Many network-based methods have been proposed for drug repositioning and most of them apply random walk on a heterogeneous network consisted with disease and drug nodes. However, these methods generally adopt the same walk-length for all nodes, and ignore the different contributions of different nodes.ResultsIn this study, we propose a drug repositioning approach based on individual bi-random walks (DR-IBRW) on the heterogeneous network. DR-IBRW firstly quantifies the individual work-length of random walks for each node based on the network topology and knowledge that similar drugs tend to be associated with similar diseases. To account for the inner structural difference of the heterogeneous network, it performs bi-random walks with the quantified walk-lengths, and thus to identify new indications for approved drugs. Empirical study on public datasets shows that DR-IBRW achieves a much better drug repositioning performance than other related competitive methods.ConclusionsUsing individual random walk-lengths for different nodes of heterogeneous network indeed boosts the repositioning performance. DR-IBRW can be easily generalized to prioritize links between nodes of a network.

Centrality ranking in multiplex networks using topologically biased random walks

Identifying influential nodes in complex networks is crucial for applications like information diffusion, rumor propagation, and epidemic spreading. Existing methods often assume uniform importance among neighbor nodes or fail to account for the structural and semantic diversity of relationships in heterogeneous networks. This leads to suboptimal identification of influential nodes, particularly in complex, attribute-rich networks. To address these shortcomings, a meta-path embedding-based graph attention network model is proposed, namely MEGA. MEGA is utilized to identify influential nodes in heterogeneous networks. This paper attempts to identify the influential set of author nodes in academic networks, defined as the set of nodes whose influence maximizes the spread of information. MEGA is described as follows. Firstly, the required meta-paths are defined, and node embeddings are obtained using a meta-path embedding method. Then, the subgraph networks corresponding to each meta-path network are generated based on random walks. Further, the graph attention network is employed to assign different weight values to neighbor nodes within the subgraph networks, indicating the varying influence of nodes on their neighbor nodes. Finally, the influence of nodes in the subgraph network is calculated using entropy, and the influence values from all subgraphs are aggregated to obtain the final influence value of each node. The effectiveness of MEGA is validated using analysis metrics from the independent cascade model. Experimental results demonstrate MEGA exceeds six baseline methods in terms of the influence spread across three datasets.

A recently proposed self-assessment questionnaire aimed at diagnosing sensitive skin provided promising results in a relatively small population. The main objectives were: (i) to assess the reliability of the aforementioned questionnaire in a larger population and verify the cut-off score previously found to predict skin sensitivity (defined as positivity to LAST, lactic acid stinging test) and (ii) to define a formula that yields the probability of a positive LAST result. Adult volunteers were included in this observational, cross-sectional, extended study. Both LAST-positive subjects, who were considered as having sensitive skin ('patients') and negative ones ('controls') completed the questionnaire, which concerned sensitivity to possible triggers of unpleasant skin sensations in real life. A cumulative score (questionnaire-based skin sensitivity score, 0-10) was calculated from the sum of all items. Three hundred and sixty-four subjects were enrolled, 214 patients and 150 controls. The mean questionnaire-based skin sensitivity score was significantly higher among patients than controls. Using two different methods, cut-off values of 4 and 5 were defined for the identification of LAST-positive subjects, with 76.6% and 72.8% accuracy, respectively. Scores below 4 or above 5 showed a high (80% or better) negative or positive predictive value, respectively. The coefficients found that in multivariate analysis for each questionnaire item, gender and age allowed us to calculate the probability of LAST positivity with higher precision taking into account the 'relative weight' of each factor. With small variations in the results, the self-assessment questionnaire confirmed its reliability for diagnosing sensitive skin in clinical practice.

Link prediction is a key tool for studying the structure and evolution mechanism of complex networks. Recommending new friend relationships through accurate link prediction is one of the important factors in the evolution, development, and popularization of social networks. At present, scholars have proposed many link prediction algorithms based on the similarity of local information and random walks. These algorithms help identify actual missing and false links in various networks. However, the prediction results significantly differ in networks with various structures, and the prediction accuracy is low. This study proposes a method for improving the accuracy of link prediction. Before link prediction, k‐shell decomposition method is used to layer the network, and the nodes that are in 1‐shell and the nodes that are not linked to the high‐shell in the 2‐shell are deleted. The experiments on four real network datasets verify the effectiveness of the proposed method.

Thermal issues are critical in 2.5D/3D IC design, and liquid cooling provides an effective solution for heat dissipation. Widely used compact thermal models (CTMs) convert chips into circuit networks for fast thermal simulations. However, current matrix-solving acceleration methods for CTM-derived circuits are inadequate for high-speed iterative transient thermal analysis of large-scale liquid-cooled 2.5D/3D ICs during design optimization. In contrast, the random walk method can provide fast solutions for local nodes in large-scale circuit networks, but it is not applicable to the circuit networks of the CTMs with liquid cooling. In this paper, we propose LCTMwalk, a novel GPU-accelerated random walk method for transient thermal analysis of liquid-cooled 2.5D/3D ICs. To enable random walks on the liquid-cooled CTM-derived circuit network, we replace the voltage-controlled current source model with the diode model. Additionally, we improve the transient analysis by using a time-backward random walk with time-domain path reuse, accelerating the solution of temperature at local circuit nodes. Experimental results show LCTMwalk can solve million-scale cases in only 500 ms, and achieves a 14-22× speedup compared to the state-of-the-art alternating direction implicit (ADI) method with GPU. Besides, LCTMwalk exhibits good generalizability and can be applied to various 2.5D/3D IC structures with high accuracy (error<1 K compared to 3D-ICE).

Scientific publications are often associated with a set of keywords to describe their content. Automating the process of keyword extraction and assignment could be useful in indexing electronic documents and building digital libraries. In this paper we propose a new approach to biomedical Concept Extraction (CE) using semantic features of concept graphs. We represent full-text documents by graphs and map biomedical terms to predefined ontology concepts. We adopt concept relation weights to improve the ranking process of potential key concepts. We perform both objective and human-based subjective evaluations. The results show that using relation weights significantly improves the performance of CE. The results also highlight the subjectivity of the CE procedure as well as of its evaluation.

In this paper, we present a framework to compare the differences in the occupation probabilities of two random walk processes, which can be generated by modifications of the network or the transition probabilities between the nodes of the same network. We explore a dissimilarity measure defined in terms of the eigenvalues and eigenvectors of the normalized Laplacian of each process. This formalism is implemented to examine differences in the diffusive dynamics described by circulant matrices, the effect of new edges, and the rewiring in networks as well as to evaluate divergences in the transport in degree-biased random walks and random walks with stochastic reset. Our results provide a general tool to compare dynamical processes on networks considering the evolution of states and capturing the complexity of these structures.

We introduce a refined way to diffusely explore complex networks with stochastic resetting where the resetting site is derived from node centrality measures. This approach differs from previous ones, since it not only allows the random walker with a certain probability to jump from the current node to a deliberately chosen resetting node, rather it enables the walker to jump to the node that can reach all other nodes faster. Following this strategy, we consider the resetting site to be the geometric center, the node that minimizes the average travel time to all the other nodes. Using the established Markov chain theory, we calculate the Global Mean First Passage Time (GMFPT) to determine the search performance of the random walk with resetting for different resetting node candidates individually. Furthermore, we compare which nodes are better resetting node sites by comparing the GMFPT for each node. We study this approach for different topologies of generic and real-life networks. We show that, for directed networks extracted for real-life relationships, this centrality focused resetting can improve the search to a greater extent than for the generated undirected networks. This resetting to the center advocated here can minimize the average travel time to all other nodes in real networks as well. We also present a relationship between the longest shortest path (the diameter), the average node degree and the GMFPT when the starting node is the center. We show that, for undirected scale-free networks, stochastic resetting is effective only for networks that are extremely sparse with tree-like structures as they have larger diameters and smaller average node degrees. For directed networks, the resetting is beneficial even for networks that have loops. The numerical results are confirmed by analytic solutions. Our study demonstrates that the proposed random walk approach with resetting based on centrality measures reduces the memoryless search time for targets in the examined network topologies.

The ability to deploy such systems throughout a region, or perhaps a nation, will increase as IoT infrastructure develops. Although Internet of Things (IoT) devices are capable of offering intelligent services, the massive amounts of data they gather and analyze pose a severe security threat. Researchers have spent a lot of work building sophisticated NIDS to prevent the exploitation of IoT data in smart applications. However, present methods may be vulnerable to attack, particularly new attacks, due to a lack of sufficient and evenly distributed attack data used to train the detection strategy. This study introduces a original hierarchical adversarial attack (HAA) generation strategy, which in turn targets the convolutional neural network-based intrusion detection in IoT devices with a limited budget, and therefore realizes the level-aware black-box adversarial attack approach. To generate adversarial instances with little disturbance, an intelligent approach is developed that relies on a saliency map methodology. To do this, a shadow CNN framework has been constructed. Using a random walk having restart based hierarchical node selection approach, a subset of the targeted IoT network's nodes with the lowest structural strength are selected. The proposed HAA generation technique is evaluated in comparison to three reference methods using the publicly available data set UNSW-SOSR2019. In a comparison of the two most sophisticated convolutional neural network (CNN) designs, GCN and JK-Net, it was shown that the classification accuracy for NIDS in IoT settings may be reduced by more than 30%.

BackgroundWe recently published in BMC Systems Biology an approach for calculating the perturbation amplitudes of causal network models by integrating gene differential expression data. This approach relies on the process of score aggregation, which combines the perturbations at the level of the individual network nodes into a global measure that quantifies the perturbation of the network as a whole. Such "bottom-up" aggregation relates the changes in molecular entities measured by omics technologies to systems-level phenotypes. However, the aggregation method we used is limited to a specific class of causal network models called "causally consistent", which is equivalent to the notion of balance of a signed graph used in graph theory. As a consequence of this limitation, our aggregation method cannot be used in the many relevant cases involving "causally inconsistent" network models such as those containing negative feedbacks.FindingsIn this note, we propose an algorithm called "sampling of spanning trees" (SST) that extends our published aggregation method to causally inconsistent network models by replacing the signed relationships between the network nodes by an appropriate continuous measure. The SST algorithm is based on spanning trees, which are a particular class of subgraphs used in graph theory, and on a sampling procedure leveraging the properties of specific random walks on the graph. This algorithm is applied to several cases of biological interest.ConclusionsThe SST algorithm provides a practical means of aggregating nodal values over causally inconsistent network models based on solid mathematical foundations. We showed its utility in systems biology, where the nodal values can be perturbation amplitudes of protein activities or gene differential expressions, while the networks can be models of cellular signaling or expression regulation. Since the SST algorithm is based on general graph-theoretical considerations, it is scalable to arbitrary graph sizes and can potentially be used for performing quantitative analyses in any context involving signed graphs.Electronic supplementary materialThe online version of this article (doi:10.1186/1756-0500-7-516) contains supplementary material, which is available to authorized users.

For a network meta-analysis, an interlinked network of nodes representing competing treatments is needed. It is often challenging to define the nodes as these typically refer to similar but rarely identical interventions. The objectives of this paper are as follows: (i) to present a series of network meta-analysis models that account for variation in the definition of the nodes and (ii) to exemplify the models where variation in the treatment definitions relates to the dose. Starting from the model that assumes each node has a 'fixed' definition, we gradually introduce terms to explain variability by assuming that each node has several subnodes that relate to different doses. The effects of subnodes are considered monotonic, linked with a 'random walk', random but exchangeable, or have a linear pattern around the treatment mean effect. Each model can be combined with different assumptions for the consistency of effects and might impact on the ranking of the treatments. Goodness of fit, heterogeneity and inconsistency were assessed. The models are illustrated in a star network for the effectiveness of fluoride toothpaste and in a full network comparing agents for multiple sclerosis. The fit and parsimony measures indicate that in the fluoride network the impact of the dose subnodes is important whereas in the multiple sclerosis network the model without subnodes is the most appropriate. The proposed approach can be a useful exploratory tool to explain sources of heterogeneity and inconsistency when there is doubt whether similar interventions should be grouped under the same node.

Recent studies have shown that adding explicit social trust information to social recommendation significantly improves the prediction accuracy of ratings, but it is difficult to obtain a clear trust data among users in real life. Scholars have studied and proposed some trust measure methods to calculate and predict the interaction and trust between users. In this article, a method of social trust relationship extraction based on hellinger distance is proposed, and user similarity is calculated by describing the f-divergence of one side node in user-item bipartite networks. Then, a new matrix factorization model based on implicit social relationship is proposed by adding the extracted implicit social relations into the improved matrix factorization. The experimental results support that the effect of using implicit social trust to recommend is almost the same as that of using actual explicit user trust ratings, and when the explicit trust data cannot be extracted, our method has a better effect than the other traditional algorithms.

Similar diseases are often caused by their similar molecular origins, such as disease-related protein-coding genes (PCGs). And nowadays, the function of PCGs has been widely studied on a gene function network, where each node represents a gene and each edge indicates an interaction between pair-wise genes. Therefore, functional interaction between disease-related PCGs should be exploited to measure disease similarity. Actually, functional interaction of pair-wise PCGs has been introduced to calculate disease similarity recently. However, existing method ignores that genes could also be associated based on intermediate nodes in the gene functional network. Here, in this article, we proposed a novel method, InfDisSim, to infer disease similarity. InfDisSim models the information flow to the network based on random walk with damping, in which the entire network could be fully utilized. The performance of InfDisSim was evaluated by a benchmark set of similar disease pairs. The area under the receiver operating characteristic curve (AUC) was calculated to evaluate the performance. As a result, InfDisSim achieves a very high AUC (0.9786), which shows it performs well. Furthermore, based on the disease similarity computed by the infDisSim, we re-validated that similar diseases tend to have common therapeutic drugs (Pearson correlation γ2=0.1315, p=2.2e−16). Finally, InfDisSim disease similarity was exploited to construct a lncRNA similarity network (LSN), which was further applied to predict potential associations between diseases and lncRNAs. High AUC (0.9893) based on leave-one-out cross validation shows the LSN is very suitable for identifying novel disease-related lncRNAs.