Parallel architectures with regular structure: A case study in modelling using stochastic well-formed coloured Petri nets

Social network analysis has typically concerned analysis of one type of tie connecting nodes of the same type. It has however been recognised that people are connected through multiple types of ties and that people in addition are affiliated with multiple types of non-people nodes. Exponential random graph models (ERGM) is a family of statistical models for social networks that at this point allows for a number of different types of network data, including one-mode networks, bipartite networks, multiplex data, as well as multilevel network data. Multilevel networks have been proposed as a joint representation of associations between multiple types of entities or nodes, such as people and organization, where two types of nodes gives rise to three distinct types of ties. The typical roster data collection method may be impractical or infeasible when the node sets are hard to detect or define or because of the cognitive demands on respondents. Multilevel multilayered networks allow us to consider a multitude of different sources of data and to sample on different types of nodes and relations. We consider modelling multilevel multilayered networks using exponential random graph models and extend a recently developed Bayesian data-augmentation scheme to allow for partially missing data. We illustrate the proposed inference procedures for the case of multilevel snowball sampling and sampling with error based on the Noordin Top network.

RA-HGNN: Attribute completion of heterogeneous graph neural networks based on residual attention mechanism

Robustness of interdependent supply chain networks against both functional and structural cascading failures

The high-level decision-making is time-consuming in a large, complex circumstance. The decision-maker needs to consider many factors whatever they are in same domain or not. Influence diagram (ID) is an effective tool to help people make a strategic decision. In a large influence diagram, the high-level decision-maker doesn't need to know the details, and he only needs to know the overall effects of the decisions made in each domain. Starting from this real application, it is necessary to aggregate the nodes in a domain into one block in an influence diagram. In this paper, we are to discuss the aggregation of nodes in influence diagrams. In order to aggregate nodes, we have to partition the influence diagram into blocks first. Each block is composed of nodes which contact each other strongly and contact other nodes loosely. Since the influence diagram has different types of nodes, we have to partition these types of nodes respectively. Based on the relationships among nodes, we aggregate each block into a new node. Further, we combine the blocks with different types of nodes and get a new influence diagram. Preliminary experiments show the feasibility of our proposed methods.

Evolving heterogeneous networks, which contain different types of nodes and links that change over time, appear in many domains including protein---protein interactions, scientific collaborations, telecommunications. In this paper, we aim to discover temporal information from a heterogenous evolving network in order to improve node classification. We propose a framework, Genetic Algorithm enhanced Time Varying Relational Classifier for evolving Heterogeneous Networks (GA-TVRC-Het), to extract the effects of different relationship types in different time periods in the past. These effects are discovered adaptively by utilizing genetic algorithms. A relational classifier is extended as the classification method in order to be able to work with different types of nodes. The proposed framework is tested on two real world data sets. It is shown that using the optimal time effect improves the classification performance to a large extent. It is observed that the optimal time effect does not necessarily follow a certain functional trend, for example linear or exponential decay in time. Another observation is that the optimal time effect may be different for each type of interaction. Both observations reveal the reason why GA-TVRC-Het outperforms methods that rely on a predefined form of time effect or the same time effect for each link type.

A relation-aware heterogeneous graph convolutional network for relationship prediction

Frequency stability in modern power network from complex network viewpoint

Considering the characteristic of heterogeneous wireless sensor networks, the clustering algorithm based on data content for heterogeneous sensor network is proposed. In cluster construction stage, according to the various sensor node types in CADC algorithm, the same types of nodes is divided into the same cluster and different types of nodes can not be divided into the same cluster. In the cluster heads selection scheme, nodes with the most residual energy will be elected cluster heads first. Cluster heads with the same type further integrate into the secondary cluster. Because this algorithm makes the data aggregation to be the most optimal and reduces the transmission distance of cluster heads, it can prolong the lifetime of the whole networks.

Railway network delay evolution: A heterogeneous graph neural network approach

In 5G networks, wide-band mmWave can be used to provide extreme data rates to user equipments (UEs). However, since mmWave is coverage limited, it necessitates dense placement of base stations, which in turn can significantly increase the fiber deployment cost. One solution being considered is to replace fibers with Integrated Access and Backhaul (IAB) network, where a part of the wireless spectrum is used for connecting base stations. In an asymmetric IAB network, standard proportional fair scheduling algorithm (PF) fails to distribute resources among UEs fairly. We propose a new weighted proportional fair (WPF) scheduling algorithm to improve the fairness of UE’s achieved throughput in an IAB network. Also to mitigate congestion in IAB nodes and improve system throughput, we propose an IABaware end-to-end flow control (I-EEFC) algorithm. Through detailed analyses for both symmetric and asymmetric network topologies, we show that our proposed combined scheduling and flow control algorithm (WPF+I-EEFC) improves both fairness and system throughput.

The paper describes the design of a simulator capable of investigating a number of ATM flow-control algorithms. The event driven simulator contains a minimum number of component types-queues, connections and routes. The first two are the only participants generating events. The major data elements are packets which logically can contain any number of cells but the simulator provides for packet transport which is statistically similar to that occurring in individual cell transport through a network. The simulator code is written in C++. Two of its mechanisms, overloaded data structures and dynamic assignment of code using pointers to functions allow the simulator to contain any number of flow control algorithms and to select the algorithm to be used. Because of the unique character of the basic simulator components, relational database structures are used for case input and run output. Run output post processing is necessary because network flow-control problems are usually related to particular components when selecting the flow control algorithm and/or parameters to best meet a system's needs. Simulator run time is excellent, with a 7 queue, 5 connection network taking 2-3 second clock time for a 0.5 second run time. Scaling to a component count by an order of magnitude should be very feasible. To demonstrate the simulator's versatility, we provide a comparison of three very different flow control schemes. The first uses fair-share assignment for individual connections at each queue. The second, a threshold queue algorithm does not control individual connections, only the bulk queue load. The third scheme is an adaptive rate-based algorithm. It uses no control at the queue-only measure transport delay at the receiver.

The mobility factor of the nodes in a mobile Ad Hoc networks (MANET) changes the network topology leading to changes in size of the network. As the topology changes, link failure between the nodes takes place due to several reasons like channel interference and dynamic obstacles that give rise to severe performance degradation. Scalability and link failure recovery are the main criteria that will determine the performance of the network in terms of quality of service (QoS). While having different type of nodes, both mobile and immobile, that have several performance severity related to data transfer which is similar to the one in industrial parameter reporting and data logging, there exists a need to overcome performance related issues such as relatively faster and secured data transfer within the set of nodes. We developed the Quantum based Routing protocol (QRP) associated with Local Link Failure Recovery Algorithm (LLFR). The QRP is a routing protocol that uses DSR and AODV as an underlying protocol to improve the QoS in scalable wireless network. The QRP and LLFR establish link failure recovery spontaneously at the point of link breakage. The performance parameters like; packet delivery ratio, throughput, average end to end delay and routing overhead of the routing protocol QRP with LLFR is analyzed using NS2 simulator. Keywords: Ad Hoc Network, Link Failure, Routing Protocols, Scalability, Quality of Service.

Topology‐aware tensor decomposition for meta‐graph learning

Application of multivariate statistics and 3D visualization analysis in tacit knowledge diffusion map

The urban traffic system is one of the most critical infrastructures. It can be abstracted into a complex network composed of strongly connected groups. Understanding the composition of such networks requires a decomposition of the network into modules, taking into account the way in which network links are distributed among the nodes, which is not an easy task. In such networks, there generally exists for each set of highly interlinked nodes little connectivity with the next set of highly interlinked nodes. This property of nodes makes community detection an important approach for analyzing the structure of complex networks. Today the emerging of big data traffic and the growing availability of smart connected devices [1] such as GPS and other sensors contribute a lot to defining the daily travel activity of individuals and giving a description for the spatial structure of travel modes in a city. In this paper, we propose a solution for exploiting urban big data traffic collected from smart devices, in order to make it available for high level applications. This will be achieved by identifying the underlying structures of the vehicles, the travellers and the stations networks and detecting the multi-layer communities in these networks. We present a global architecture for an intelligent transport network with some details of each its components. The main contribution of the proposed architecture is the Community Analysis component, which allows us to analyse communities with different types of nodes: travelers with their preferences, public transport stations and public vehicles. Then using the community detection method and the network analysis method, we present one application of the proposed system that is related to public transportation optimization. To take into account travelers' preferences, we propose a traveler's profile model and we show how it can help to detect travelers' communities.