Number Of Subnetworks Research Articles

Stability analysis of fractional order memristor synapse-coupled hopfield neural network with ring structure.

A memristor is a nonlinear two-terminal electrical element that incorporates memory features and nanoscale properties, enabling us to design very high-density artificial neural networks. To enhance the memory property, we should use mathematical frameworks like fractional calculus, which is capable of doing so. Here, we first present a fractional-order memristor synapse-coupling Hopfield neural network on two neurons and then extend the model to a neural network with a ring structure that consists of n sub-network neurons, increasing the synchronization in the network. Necessary and sufficient conditions for the stability of equilibrium points are investigated, highlighting the dependency of the stability on the fractional-order value and the number of neurons. Numerical simulations and bifurcation analysis, along with Lyapunov exponents, are given in the two-neuron case that substantiates the theoretical findings, suggesting possible routes towards chaos when the fractional order of the system increases. In the n-neuron case also, it is revealed that the stability depends on the structure and number of sub-networks.

A Two-Stage Decomposition Approach for AC Optimal Power Flow

The alternating current optimal power flow (AC-OPF) problem is critical to power system operations and planning, but it is generally hard to solve due to its nonconvex and large-scale nature. This paper proposes a scalable decomposition approach in which the power network is decomposed into a master network and a number of subnetworks, where each network has its own AC-OPF subproblem. This formulates a two-stage optimization problem and requires only a small amount of communication between the master and subnetworks. The key contribution is a smoothing technique that renders the response of a subnetwork differentiable with respect to the input from the master problem, utilizing properties of the barrier problem formulation that naturally arises when subproblems are solved by a primal-dual interior-point algorithm. Consequently, existing efficient nonlinear programming solvers can be used for both the master problem and the subproblems. The advantage of this framework is that speedup can be obtained by processing the subnetworks in parallel, and it has convergence guarantees under reasonable assumptions. The formulation is readily extended to instances with stochastic subnetwork loads. Numerical results show favorable performance and illustrate the scalability of the algorithm which is able to solve instances with more than 11 million buses.

A taxonomy of international manufacturing networks

Manufacturing firms with multiple product groups do not need to involve all factories in the production of all product groups. Some factories may specialize on a small set of products, while others participate in the manufacturing of a broader set of products. However, current theories on international manufacturing networks do not explain in detail how organizations design international manufacturing networks for different products or product groups involving different sets of factories. This research investigates 20 product group networks at five global manufacturing firms. We distinguish between three types of factories: component manufacturing factories, assembly factories, and integrated factories (having both component manufacturing and assembly). Furthermore, we identify four network types: linear, divergent, convergent, and mixed structures. These four types exhibit distinctly different characteristics in terms of key characteristics, factory roles, product types, process types, market types, sourcing, and key managerial challenges. Most networks are relatively small – on an average consisting of four factories and some contain a number of subnetworks that are self-sufficient in terms of material flow and serve separate market regions. We identify two new types of factory roles related to component manufacturing competences, which we call ‘strategic feeder’ and ‘full lead’.

A study of Female Sex worker’s sub-network on STI treatment seeking behavior in Chennai City: A Social Network Approach

Background: Female Sex workers (FSW) are having high-risk sexual behavior with multiple partners and the periodical sexually transmitted infection screening is determined by various factors. Aims & Objectives: To explore the presence of sub-networks of FSWs and to identify the trends in information flow on Sexually Transmitted Infection (STI) treatment seeking behavior by using social network analysis (SNA). Material & Methods: This was a community-based cross-sectional study and exponential snowball sampling method is used to identify 420 FSWs. Structured questionnaire was used to measure information flow and sub-network characteristics of social relationship, Information & Treatment seeking for STI and Sexual practice. Results: Study has listed 94 sub-networks. Here, Sexual practice (Sub-group 28) has highest number of subnetworks compared to other three subnetworks of Social Relationship (Sub-group 25), information on STI (Sub-group 23) and Treatment seeking for STI (Sub-group18). Similarly, other Centrality measures like Triad and Dyad are higher (Each 4 sub-group) in sexual practice domain and this is absent in rest of the three domains. Further, treatment seeking for STI has low Clique (Sub-group 17) compared to other domains. This shows high information flow is present in Sexual Practice compare to treatment seeking for STI (CC-0.986, P-0.000). Among the sub typology home based and street based FSWs are well connected, well informed and most influential than other sub-typology. Conclusion: SNA reaches more number of HRGs with less resource. Therefore, SNA could be cost effective to mapping the sub-network and visualize the information path for STI treatment seeking behavior

Stochastic network link transmission model

This article considers the stochastic modeling of vehicular network flows, including the analytical approximation of joint queue-length distributions. The article presents two main methodological contributions. First, it proposes a tractable network model for finite space capacity Markovian queueing networks. This methodology decomposes a general topology queueing network into a set of overlapping subnetworks and approximates the transient joint queue-length distribution of each subnetwork. The subnetwork overlap allows to approximate stochastic dependencies across multiple subnetworks with a complexity that is linear in the number of subnetworks. Additionally, the network model maintains mutually consistent overlapping subnetwork distributions. Second, a stochastic network link transmission model (SLTM) is formulated that builds on the proposed queueing network decomposition and on the stochastic single-link model of Osorio and Flötteröd (2015). The SLTM represents each direction of a road and each road intersection as one queueing subnetwork. Three experiments are presented. First, the analytical approximations of the queueing-theoretical model are validated against simulation-based estimates. An experiment with intricate traffic dynamics and multi-modal joint distributions is studied. The analytical model captures most dependency structure and approximates well the simulated network dynamics and joint distributions. Even for the considered simple network, which consists of only eight links, the proposed subnetwork decomposition yields significant gains in computational efficiency: It uses less than 0.0025% of the memory that is required by the use of a full network model. Second and third, the proposed SLTM is illustrated with a linear test network adopted from the literature and a more general topology network containing a diverge node and a merge node. Time-dependent probabilistic performance measures (occupancy uncertainty bands, spillback probabilities) are presented and discussed.

Implementation and Analysis of Adaptive Packet Throttling in Mesh NoCs

Network-on-Chip (NoC) has emerged as the preferred communication framework for multi-core systems. Packet throttling is a cost effective technique which simply delays packet injection into the network. However, deciding on when to throttle and how much to throttle are key design challenges of any throttling technique. Existing techniques use local throttling decisions by a central controller. This paper overcome the issues related with the existing works by partitioning the network into number of sub-networks, each with a zonal controller. The experimental results with real traffic workloads show substantial reduction in average packet latency when compared with state of the art packet throttling techniques.

Partitioned modal extraction applied to power system wide-area measurements using empirical orthogonal functions

This paper presents a numerical algorithm in order to identify and extract the inter-area modal inestability presented in large interconnected power systems, which significantly decreases the computational burden and gives a distributed application in the analysis of collected data from wide-area measurement systems (WAMS) using the empirical orthogonal functions. The proposed algorithm is based on a spatio-temporal statistical identification method using both empirical orthogonal functions (EOF) analysis and partitioned network data approximation. Firstly, identification process is carried out separating the power system network into a number of subnetworks either geographically distributed or by containing the same load types. Then, EOF analysis is applied to the local data array which contains the space-time response for each subnetwork where modal extraction is determined. Afterward, the derived modal approximation sets are used to compute global predominant modes by correlation of its temporal coefficients, which yields to a refined global approximation of local and inter-area oscillations, and a model-order reduction removing spurious modes. An illustrative example involving a transmission system is presented in order to validate the proposed methodology.

Sequential patterns of spikes and scale-invariance in modular networks

It has been reported that there are consistent sequential patterns of spikes after the transitions to the up state during slow wave sleep[1]. The up states may be characterized by critical dynamics[2] for which the avalanche sizes distribution is scale-invariant[3]. In order to understand the mechanism of the sequential patterns, it may thus be necessary to study the fine structure of avalanche transmission between multiple neuronal ensembles at criticality. We have developed an analytical model of avalanche dynamics in modular networks. The univariate distribution of avalanche sizes[4] can be extended to a joint probability distribution Pw({Lu}u∈S) describing the probability that Lu neurons are active in each sub-network u during avalanches that start from the sub-network w with u∈S={1,...,M} and M the number of sub-networks. The mean temporal profile[3] of the avalanches in each sub-network u can then be detailed (see Figure ​Figure11 (A1) and (B1)) and are reminiscent of the temporal patterns observed experimentally. The sequential patterns depend on the average connection strengths between the sub-networks. At criticality, the cumulated temporal profiles can be collapsed using standard rescaling of the axes and yield a single universal scaling function[3] (see Figure ​Figure11 (A2) and (A3)). After rescaling, the mean spiking times in each sub-network are functions of the durations of the avalanches. In Figure ​Figure11 (B3), the intervals between successive mean spiking times - are proportionally shorter for longer avalanches near criticality. Figure 1 (A1) The average number of neurons spiking at the time-step t in the sub-networks u=1, 2, and 3 are shown red, blue, and green, respectively, for avalanches of duration τ=15 starting from the sub-network 1. (B1) ...

EXtended Torus routing algorithm for networks‐on‐chip: a routing algorithm for dynamically reconfigurable networks‐on‐chip

This paper presents a novel routing algorithm called eXtended Torus routing algorithm for networks-on-chip (XTRANC) which supports topologyies based on a variable number and size of inner-torus building blocks. The inner-tori partition a traditional mesh network into an arbitrary number of sub-networks to increase the mesh performance. The sub-networks can generate non-regular global topologies which are also supported by the XTRANC algorithm. XTRANC is especially suitable for dynamically reconfigurable networks mapped to commercial FPGAs in which additional links are added to the mesh topology at run-time to reduce congestion depending on application behaviour and resource availability. XTRANC allows the insertion of links as requested by different parts of the application without centralized control and this research shows that despite this dynamic behaviour the routing algorithm remains deadlock free.

Delay-induced synchronization transitions in modular scale-free neuronal networks with hybrid electrical and chemical synapses

We study the dependence of synchronization transitions in modular networks of bursting neurons with hybrid electrical–chemical synapses on the information transmission delay and the probability of electrical synapses. The modular network is composed of subnetworks (clusters); each of them presents the scale-free property. It is shown that, irrespective of the probability of electrical synapses, the time delay can always induce synchronization transitions in modular neuronal networks. Regions of synchronization and non-synchronization appear intermittently as the delay increases. In particular, all these transitions to burst synchronization occur approximately at integer multiples of oscillatory period of individual neurons. In addition, for larger probability of electrical synapses, the intermittent synchronization transition is more profound, due to the stronger synchronization capability of electrical synapses compared with chemical ones. Furthermore, the transition to synchronous bursting can also be induced by the variation of modular network parameters, that is, the coupling strength between neurons, the interconnection probability between different subnetworks, as well as the number of subnetworks. Particularly, we find that a modular neuronal network is harder to get global synchronization when constituting neurons are dispersed over more clusters. On the other hand, chemical and electrical synapses can perform complementary roles in the synchronization of hybrid modular neuronal networks: the larger the electrical synapse strength is the smaller the chemical synapse strength needed to achieve burst synchronization.

The effects of time delay on the synchronization transitions in a modular neuronal network with hybrid synapses

Delay-induced synchronization transitions are studied in a modular neuronal network of small-world subnetworks with hybrid synapses in this paper. Numerical results show that the spatiotemporal synchronization transitions in a modular neuronal network not only depend on the information transmission delay, but also can be induced by the variations of the probability of inhibitory synapses and the number of subnetworks in the modular networks. In the hybrid modular network, the information transmission delay is shown to be significant, which can either promote or destroy synchronization of neuronal activity. In particular, the increasing delays can induce the intermittent appearance of regions of synchronization and non-synchronization. Interestingly, it is found that intermittent synchronization transition is relatively profound for smaller and larger probability of inhibitory synapses, while synchronization transition seems less profound for the moderate probability of inhibitory synapses. In addition, if only the delay is appropriate, there exists a suitable modular network topology structure enhancing the synchronized neuronal activity.

A Clustering Protocol for Wireless Sensor Networks Based on Energy Potential Field

It is the core issue of researching that how to prolong the lifetime of wireless sensor network. The purpose of this paper is to illustrate a clustering protocol LEACH-PF, which is a multihop routing algorithm with energy potential field of divided clusters. In LEACH-PF, the network is divided into a number of subnetworks and each subnetwork has a cluster head. These clusters construct an intercluster routing tree according to the potential difference of different equipotential fields. The other member nodes of the subnetworks communicate with their cluster head directly, so as to complete regional coverage. The results of simulation show that LEACH-PF can reduce energy consumption of the network effectively and prolong the network lifetime.

Methods for division of road traffic network for distributed simulation performed on heterogeneous clusters

The computer simulation of road traffic is an important tool for control and analysis of road traffic networks. Due to their requirements for computation time (especially for large road traffic networks), many simulators of the road traffic has been adapted for distributed computing environment where combined power of multiple interconnected computers (nodes) is utilized. In this case, the road traffic network is divided into required number of sub-networks, whose simulation is then performed on particular nodes of the distributed computer. The distributed computer can be a homogenous (with nodes of the same computational power) or a heterogeneous cluster (with nodes of various powers). In this paper, we present two methods for road traffic network division for heterogeneous clusters. These methods consider the different computational powers of the particular nodes determined using a benchmark during the road traffic network division.

Stochastic resonance on a modular neuronal network of small-world subnetworks with a subthreshold pacemaker

We study the phenomenon of stochastic resonance on a modular neuronal network consisting of several small-world subnetworks with a subthreshold periodic pacemaker. Numerical results show that the correlation between the pacemaker frequency and the dynamical response of the network is resonantly dependent on the intensity of additive spatiotemporal noise. This effect of pacemaker-driven stochastic resonance of the system depends extensively on the local and the global network structure, such as the intra- and inter-coupling strengths, rewiring probability of individual small-world subnetwork, the number of links between different subnetworks, and the number of subnetworks. All these parameters play a key role in determining the ability of the network to enhance the noise-induced outreach of the localized subthreshold pacemaker, and only they bounded to a rather sharp interval of values warrant the emergence of the pronounced stochastic resonance phenomenon. Considering the rather important role of pacemakers in real-life, the presented results could have important implications for many biological processes that rely on an effective pacemaker for their proper functioning.

Vibrational resonance in excitable neuronal systems

In this paper, we investigate the effect of a high-frequency driving on the dynamical response of excitable neuronal systems to a subthreshold low-frequency signal by numerical simulation. We demonstrate the occurrence of vibrational resonance in spatially extended neuronal networks. Different network topologies from single small-world networks to modular networks of small-world subnetworks are considered. It is shown that an optimal amplitude of high-frequency driving enhances the response of neuron populations to a low-frequency signal. This effect of vibrational resonance of neuronal systems depends extensively on the network structure and parameters, such as the coupling strength between neurons, network size, and rewiring probability of single small-world networks, as well as the number of links between different subnetworks and the number of subnetworks in the modular networks. All these parameters play a key role in determining the ability of the network to enhance the outreach of the localized subthreshold low-frequency signal. Considering that two-frequency signals are ubiquity in brain dynamics, we expect the presented results could have important implications for the weak signal detection and information propagation across neuronal systems.

Burst synchronization transitions in a neuronal network of subnetworks

In this paper, the transitions of burst synchronization are explored in a neuronal network consisting of subnetworks. The studied network is composed of electrically coupled bursting Hindmarsh-Rose neurons. Numerical results show that two types of burst synchronization transitions can be induced not only by the variations of intra- and intercoupling strengths but also by changing the probability of random links between different subnetworks and the number of subnetworks. Furthermore, we find that the underlying mechanisms for these two bursting synchronization transitions are different: one is due to the change of spike numbers per burst, while the other is caused by the change of the bursting type. Considering that changes in the coupling strengths and neuronal connections are closely interlaced with brain plasticity, the presented results could have important implications for the role of the brain plasticity in some functional behavior that are associated with synchronization.

Evaluating the damage associated with intentional network disintegration

This paper presents a method for evaluating an expected damage associated with disintegrating complex networks with a given topology into isolated sub-networks (clusters) as a result of intentional attack on randomly chosen network links. The method is based on a multi-dimensional spectra approach for evaluating the probability of network disintegration into a given number of sub-networks when a fixed number of randomly chosen links is eliminated. It also uses the contest success function that evaluates destruction probability of individual links as a function of per-link attack and defense efforts. It is assumed that the defender has no information about the attacker's actions and the attacker has no information about the network structure. The method allows the analysts to compare different network topologies and to choose one with the minimal expected damage under conditions of uncertainty. Illustrative examples are presented.

Algorithms for partitioning of large routing networks

Partitioning of large networks is vital for decentralized management and control.This paper presents two algorithms called ‘Hierarchical Recursive Progression-1’ (HRP-1) and ‘Hierarchical Recursive Progression-2’ (HRP-2) for partitioning of large networks into subnetworks of limited size with very few interconnections between them. In other words, we are trying to maximize the internal nodes and minimize the external connections of the subnetworks. The restriction on the size and the external connections is obtained by comparison against a user-defined value for the size of the subnetwork and for external connections via a term called density. The density of a subnetwork is defined as the ratio of the number of external connections and the size of the subnetwork. The two algorithms presented in the paper are based on the principle of subnetwork clustering. At the start of the algorithms,the number of subnetworks is equal to the total number of nodes of the network with each subnetwork containing a single node. Later, subnetworks are merged at various runs of the algorithm to form new subnetworks using connectivity,density and size criteria. The algorithms terminate when all the subnetworks satisfy a user-defined size and density limit. The difference between the algorithms HRP-1 and HRP-2 lies in the definition of density of subnetworks and the way through which the subnetworks are grouped at consecutive iterations of the algorithm. Note that the number of nodes inside the subnetworks never violates the size limit, thereby ensuring even distribution of load on the partitions obtained. Finally, the two algorithms are compared and tested on randomly generated graphs and a part of Paris road Network.

Modeling measurement errors and missing initial values in freeway dynamic origin–destination estimation systems

Most existing dynamic origin–destination (O–D) estimation approaches are grounded on the assumption that a reliable initial O–D set is available and traffic volume data from detectors are accurate. However, in most traffic systems, both types of critical information are either not available or subjected to some level of measurement errors such as traffic counts and speed measurement from sensors. To contend with those critical issues, this study presents two robust algorithms, one for estimation of an initial O–D set and the other for tackling the input measurement errors with an extended estimation algorithm. The core concept of the initial O–D estimation algorithm is to decompose the target network in a number of sub-networks based on proposed rules, and then execute the estimation of the initial O–D set iteratively with the observable information at the first time interval. To contend with the inevitable detector measurement error, this study proposes an interval-based estimation algorithm that converts each model input data as an interval with its boundaries being set based on some prior knowledge. The performance of both proposed algorithms has been tested with a simulated system, the I-95 freeway corridor between I-495 and I-695, and the results are quite promising.

A multimodal neural network with single-state predictions for protein secondary structure

Prediction of protein secondary structure is considered to be an important step toward elucidating the three-dimensional structure and function of proteins. We have developed a multimodal neural network (MNN) to predict protein secondary structure. The MNN is composed of several subclassifiers for single-state predictions using neural networks and a decision neural network (DNN). Each subclassifier employs a number of subnetworks to predict the single-state of the secondary structure individually and produces the final results by majority decision. The DNN uses a three-layer neural network to produce the final overall prediction from the outputs of the single-state predictions. The MNN gives an overall accuracy of 71.1% with corresponding Matthews correlation coefficients of CH = 0.62 and CE = 0.53. The prediction test is based on a database of 126 nonhomologous protein sequences.