Types Of Network Models Research Articles

Development of a Nonlinear Soft-Sensor Using a GMDH Network for a Refinery Crude Distillation Tower

SUMMARY In atmospheric distillation processes, stabilization of processes is required in order to optimize the crude-oil composition that corresponds to product market conditions. However, the process control systems sometimes fall into unstable states in the case where unexpected disturbances are introduced, and these unusual phenomena have had an undesirable effect on certain products. Furthermore, a useful chemical engineering model has not yet been established for these phenomena. This remains a serious problem in atmospheric distillation processes. This paper describes a new modeling scheme to predict unusual phenomena in the atmospheric distillation process using a GMDH (Group Method of Data Handling) network, which is a type of network model. According to the GMDH network, the model structure can be determined systematically. However, the method of least squares has been commonly utilized in determining weight coefficients (model parameters). Estimation accuracy is not entirely essential, because the sum of the squared errors between the measured values and estimates is evaluated. Therefore, instead of evaluating the sum of the squared errors, the sum of the absolute values of the errors is introduced and the Levenberg–Marquardt method is employed in order to determine the model parameters. The effectiveness of the proposed method is evaluated by foaming prediction in the crude oil switching operation in the atmospheric distillation process.

Modelling Framework and the Quantitative Analysis of Distributed Energy Resources in Future Distribution Networks

Abstract There has been a large body of statements claiming that the large-scale deployment of Distributed Energy Resources (DERs) could eventually reshape the future distribution grid operation in numerous ways. Thus, it is necessary to introduce a framework to measure to what extent the power system operation will be changed by various parameters of DERs. This article proposed a modelling framework for an overview analysis on the correlation between DERs. Furthermore, to validate the framework, the authors described the reference models of different categories of DERs with their unique characteristics, comprising distributed generation, active demand and electric vehicles. Subsequently, quantitative analysis was made on the basis of the current and envisioned DER deployment scenarios proposed for Sweden. Simulations are performed in two typical distribution network models for four seasons. The simulation results show that in general the DER deployment brings in the possibilities to reduce the power losses and voltage drops by compensating power from the local generation and optimizing the local load profiles.

Cloud MapReduce for Monte Carlo bootstrap applied to Metabolic Flux Analysis

The MapReduce architectural pattern popularized by Google has successfully been utilized in several scientific applications. Up until now, MapReduce is rarely employed in the field of Systems Biology. We investigate whether a MapReduce approach utilizing on-demand resources from a Cloud is suitable to perform simulation tasks in the area of Metabolic Flux Analysis (MFA). An Amazon ElasticMapReduce Cloud implementation of the parallel, parametric Monte Carlo bootstrap in the context to 13C-MFA is presented. The seamless integration of the application into a service-oriented, BPEL-based scientific workflow framework is shown. A comparison of a straightforward MapReduce implementation using the Hadoop streaming interface on various Amazon ElasticMapReduce instance types and a single CPU core computation approach reveals a speedup of 17 on 64 Amazon cores. I/O operations on many small files within the Reduce step were identified as the limiting step. By exploiting the Hadoop Java API, making use of built-in data types and tuning problem-specific Hadoop parameters, the I/O issues could be resolved. With the revised implementation, a speedup of up to 48 could be achieved on 64 Amazon cores. To investigate the runtimes of a realistic 13C-MFA analysis, 50,000 Monte Carlo samples with a typical metabolic network model have been performed on 20 virtual nodes in 24 h and 23 min with a total cost of $384. Our work demonstrates the possibility to perform scalable Systems Biology applications using Amazon’s Cloud MapReduce service.

Sum Rate Optimization in Multiuser Cognitive Radio Networks

In multiuser cognitive radio (CR) networks, we address the problem of joint transmit beamforming (BF) and power control (PC) for secondary users (SUs) when they are allowed to transmit simultaneously with primary users (PUs). The objective is to optimize the network sum rate under the interference constraints of PUs, which is a nonconvex problem. Iterative dual subgradient (IDuSuG) algorithm is proposed by iteratively performing BF and PC to optimize the sum rate, among which minimum mean square error (MMSE) or virtual power-weighed projection (VIP2) is used to design beamformers and subgradient method is used to control the power. VIP2 algorithm is devised for the case in which the interference caused by MMSE beamformer exceeds the threshold. Moreover, channel uncertainty due to lack of cooperation is considered. A closed-form worst-case expression is derived, with which the uncertainty optimization problem is transformed into a certain one. A robust algorithm based on IDuSuG is provided by modifying updates in iterative process. Furthermore, second-order cone programming approximation (SOCPA) method is proposed as another robust algorithm. Typical network models are approximated to SOCP problems and solved by interior-point method. Finally the network sum rates for different PU and SU numbers are assessed for both certainty and uncertainty channel models by simulation.

Communication model and protocol based on multiple static sinks for supporting mobile users in wireless sensor networks

Typical communication model of wireless sensor networks consists of users, sinks, and a number of sensor nodes. The users are remote from wireless sensor networks and they gather data from the sinks via legacy networks. However, in practical sensor network applications, there are two types of users: traditional remote users and mobile users such as fire-fighters and soldiers. The mobile users may move around sensor fields and they communicate with the sinks only via the sensor networks in order to gather data like location information of victims in disaster areas. In this paper, in order to effectively support both the remote users and the mobile users, we propose a novel communication model relying on the typical sensor network model. In the model, multiple static sinks connect with legacy networks and divide a sensor field into the number of the multiple sinks. Through sharing queries and data via the legacy networks, the multiple static sinks provide high throughput through distributed data gathering and low latency through short-hops data delivery. Multiple static sinks deliver the aggregated data to the remote users via the legacy networks. In case of the mobile users, when a mobile user moves around, it receives the aggregated data from the nearest static sink. Simulation results show that the proposed model is more efficient in terms of energy consumption, data delivery ratio, and delay than the existing models.

On the distributions of Laplacian eigenvalues versus node degrees in complex networks

In this paper, the important issue of Laplacian eigenvalue distributions is investigated through theory-guided extensive numerical simulations, for four typical complex network models, namely, the ER random-graph networks, WS and NW small-world networks, and BA scale-free networks. It is found that these four types of complex networks share some common features, particularly similarities between the Laplacian eigenvalue distributions and the node degree distributions.

Semantic Priming in a Cortical Network Model

Contextual recall in humans relies on the semantic relationships between items stored in memory. These relationships can be probed by priming experiments. Such experiments have revealed a rich phenomenology on how reaction times depend on various factors such as strength and nature of associations, time intervals between stimulus presentations, and so forth. Experimental protocols on humans present striking similarities with pair association task experiments in monkeys. Electrophysiological recordings of cortical neurons in such tasks have found two types of task-related activity, "retrospective" (related to a previously shown stimulus), and "prospective" (related to a stimulus that the monkey expects to appear, due to learned association between both stimuli). Mathematical models of cortical networks allow theorists to understand the link between the physiology of single neurons and synapses, and network behavior giving rise to retrospective and/or prospective activity. Here, we show that this type of network model can account for a large variety of priming effects. Furthermore, the model allows us to interpret semantic priming differences between the two hemispheres as depending on a single association strength parameter.

Characterization of subgraph relationships and distribution in complex networks

A network can be analyzed at different topological scales, ranging from single nodes to motifs, communities, up to the complete structure. We propose a novel approach which extends from single nodes to the whole network level by considering non-overlapping subgraphs (i.e. connected components) and their interrelationships and distribution through the network. Though such subgraphs can be completely general, our methodology focuses on the cases in which the nodes of these subgraphs share some special feature, such as being critical for the proper operation of the network. The methodology of subgraph characterization involves two main aspects: (i) the generation of histograms of subgraph sizes and distances between subgraphs and (ii) a merging algorithm, developed to assess the relevance of nodes outside subgraphs by progressively merging subgraphs until the whole network is covered. The latter procedure complements the histograms by taking into account the nodes lying between subgraphs, as well as the relevance of these nodes to the overall subgraph interconnectivity. Experiments were carried out using four types of network models and five instances of real-world networks, in order to illustrate how subgraph characterization can help complementing complex network-based studies.

Universal robustness characteristic of weighted networks against cascading failure

We investigate the cascading failure on weighted complex networks by adopting a local weighted flow redistribution rule, where the weight of an edge is (k(i)k(j))theta with k(i) and k(j) being the degrees of the nodes connected by the edge. Assume that a failed edge leads only to a redistribution of the flow passing through it to its neighboring edges. We found that the weighted complex network reaches the strongest robustness level when the weight parameter theta=1, where the robustness is quantified by a transition from normal state to collapse. We determined that this is a universal phenomenon for all typical network models, such as small-world and scale-free networks. We then confirm by theoretical predictions this universal robustness characteristic observed in simulations. We furthermore explore the statistical characteristics of the avalanche size of a network, thus obtaining a power-law avalanche size distribution together with a tunable exponent by varying theta. Our findings have great generality for characterizing cascading-failure-induced disasters in nature.

A NETWORK-BASED THRESHOLD MODEL FOR THE SPREADING OF FADS IN SOCIETY AND MARKETS

We investigate the behavior of a threshold model for the spreading of fads and similar phenomena in society. The model gives the fad dynamics and is intended to be confined to an underlying network structure. We investigate the whole parameter space of the fad dynamics on three types of network models. The dynamics we discover is rich and highly dependent on the underlying network structure. For some range of the parameter space, for all types of substrate networks, there are a great variety of sizes and life-lengths of the fads — what one sees in real-world social and economical systems.

Complex networks: Small-world, scale-free and beyond

In the past few years, the discovery of small-world and scale-free properties of many natural and artificial complex networks has stimulated a great deal of interest in studying the underlying organizing principles of various complex networks, which has led to dramatic advances in this emerging and active field of research. The present article reviews some basic concepts, important progress, and significant results in the current studies of various complex networks, with emphasis on the relationship between the topology and the dynamics of such complex networks. Some fundamental properties and typical complex network models are described; and, as an example, epidemic dynamics are analyzed and discussed in some detail. Finally, the important issue of robustness versus fragility of dynamical synchronization in complex networks is introduced and discussed.

Structural classification of network models

A newly developed structural classification of network models is suggested. It is based on singling out three main groups of characteristics which define both the structure and the parameters of network models. Those groups are (1) network elements (nodes, arrows, terms’ restriction, logical links, etc.), (2) elements’ parameters (number, function, set of variants, random variable, etc.) and (3) degree of alternativity (alternative logical operations at the node’s input and output). The structural classification arranges an order in the variety of all the types of network models by using a three-dimension matrix. Each model can be characterised by a set of cells in the three-dimensional ‘house’. The classification enables not only a description of the network models, but also a forecast of new models. The latter may be discovered either by filling in ‘empty places’ in the matrix or by implementing new attributes in the groups’ characteristics.

Nonlinear modelling and prediction with feedforward and recurrent networks

In feedforward networks, signals flow in only one direction without feedback. Applications in forecasting, signal processing and control require explicit treatment of dynamics. Feedforward networks can accommodate dynamics by including past input and target values in an augmented set of inputs. A much richer dynamic representation results from also allowing for internal network feedbacks. These types of network models are called recurrent network models and are used by Jordan (1986) for controlling and learning smooth robot movements, and by Elman (1990) for learning and representing temporal structure in linguistics. In Jordan's network, past values of network output feed back into hidden units; in Elman's network, past values of hidden units feed back into themselves. The main focus of this study is to investigate the relative forecast performance of the Elman type recurrent network models in comparison to feedforward networks with deterministic and noisy data. The salient property of the Elman type recurrent network architecture is that the hidden unit activation functions (internal states) are fed back at every time step to provide an additional input. This recurrence gives the network dynamical properties which make it possible for the network to have internal memory. Exactly how this memory is represented in the internal states is not determined in advance. Instead, the network must discover the underlying temporal structure of the task and learn to encode that structure internally. The simulation results of this paper indicate that recurrent networks filter noise more successfully than feedforward networks in small as well as large samples.

Nonlinear modelling and control of electrically stimulated muscle: A local model network approach

Human muscle can be made to contract by electrical stimulation. There are many well established uses for this technique and new applications are being investigated. We are interested in two uses: restoration of function to paralysed limbs and reconfiguration of one of the back muscles to assist a failing heart. Muscle modelling has been widely studied in the past. Model types range from biophysical models, which are based on the structure and processes of actual muscles, through analogue models, to purely mathematical descriptions. The latter are derived only from the input (neural activation) and the output (mechanical response) signals. We are searching for mathematical models, which firstly can represent the complex nonlinear behaviour of muscle and secondly are controller oriented, that is that controller design based on such models is possible. We have developed a number of dynamic muscle models based on force measurements taken from isometric contracting muscles stimulated by an irregular pulse train. The results of experimental investigations have shown that a simple linear transfer-function model of muscle response can be rather inaccurate over the full operational range. The muscle response is significantly nonlinear with respect to the level of stimulation. However, the experiments with linear system identification provide results which can be used to choose structure parameters of a dynamic model. Nonlinear models based on local model networks (LMNs) are able to capture the nonlinear effects and to provide accuracy over a wide operational range. We investigate the use of an LMN with local linear models, describe the type of network model used, the training algorithm, and show some typical experimental results based on real measured data. We address the synthesis of feedback controllers for physical muscle. The simple linear transfer function model forms the basis for the design of a robust linear controller. The structure of the LMN derived previously is exploited as the basis of a non-linear control design: the local controller network. We compare the performance of both controller designs in simulation experiments.

Comparison of k-shortest paths and maximum flow routing for network facility restoration

In the development of technologies for span failure restoration, a question arises about the restoration rerouting characteristics to be specified. In theory, maximal rerouting capacity is obtained with a maximum flow (Max Flow) criterion. However, rerouting that realizes the k-successively shortest link disjoint paths (KSP) may be faster, easier, and, in distributed implementation, more robust than a distributed counterpart for Max Flow. The issue is, therefore, what the restoration capacity penalty is if KSP is used instead of Max Flow. To explore this tradeoff, the authors present a comparative study of the effectiveness of KSP versus Max Flow as an alternative rerouting criteria in the context of transport network span restoration. The comparison applies to both centrally controlled and distributed restoration systems. Study methods include exhaustive span failure experiments on a range of network models, and parametric and analytical investigations for insight into the factors resulting in KSP versus Max Flow differences. The main finding is that KSP restoration capacity is more than 99.9% of that from Max Flow in typical network models. The hypothesis is made that a generalized "trap" topology is responsible for all KSP-Max Flow capacity differences. The hypothesis is tested experimentally and used to develop analytical bounds which agree well with observed results. These findings and data are relevant to standards makers and equipment developers in specifying and engineering future restorable networks.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Learning algorithms for oscillatory networks with gap junctions and membrane currents

One of the most important problems for studying neural network models is the adjustment of parametas. Here we show how to formulate the problem as the minimization of the dilferslee betwen two limit cycles. The backprop%ation method for leaming algorithms is described .ss the application of grsdient descent to an eror fundion that computes this difference. A mathematical formulation is given that is applicable to any type of network model, and is applied to several models. For example, when learning in a network in which dl cells have a common, adjustable, bias current, the value of the bias ie adjustcd at a rate proportional to the difference between the sum of the target outputs and the sum of the actual outputs. When learning in a network of n cells where a target output is given for every cell. the learning algorithm splits into R indepndent leaming algorithms, one per cell. For networks containing gap junctions, a gap junction is modelled D conductance times the potential difference between the two adjacent cells. The requirement that a conductme g munt be positive is enionzed by replacing g by a functbn pos(g*) whose value is always positive, for example cxp(O.lg*), and deriving an algorithm that adjusts the parameta g* in place of g. When target output is specified for every cell in a network with gap junctions. the learning algorithm splits into fewer independent componeds, one for each gap-mnneaed subset of the network. The lemming algorithm for II gspmnneeted set of cells cannot be paralklized further. As a find example, a leaming algwithm is derived for a mutually inhibitory twc- cell network in which each cell has a membrane current. This generalized approach to backpropagation allows one to derive a learning algorithm for alntoat any model neural network given in tem of differmtid equations. It will be an essential tool for adjusting parmetem in small but complex network models.

Constitution of wide area telephone network using optical fiber communication

AbstractWith recent, remarkable development of optical fiber technology optical fibers are replacing conventional coaxial cables and microwave systems in the toll telephone network. This paper optimizes a telecommunication network configuration which uses large‐capacity transmission systems such as optical fibers. In a hypothetical Japanese toll telephone network, the required numbers of circuits and systems are calculated and compared for five types of network models (i.e., mesh, lattice, ring, ring‐mesh composite and star‐mesh composite types). As a result of these calculations, the 400–Mb/s system and 1.6–Gb/s system are shown to be applicable to most of the toll transmission sections. The required amounts of circuits increase in the order of lattice type, ring‐mesh composite type, star‐mesh composite type, mesh type and ring type. When the total toll originating traffic becomes about 5 × 106 erl (it is about 106 erl at present), the differences in required numbers of circuits among four types except the ring type are within several percent.