Network Structure Parameters Research Articles

Evaluation of Financial Performance of Port Enterprises Based on Radial Basis Function Neural Network

Li, J., 2020. Evaluation of financial performance of port enterprises based on radial basis function neural network. In: Gong, D.; Zhang, M., and Liu, R. (eds.), Advances in Coastal Research: Engineering, Industry, Economy, and Sustainable Development. Journal of Coastal Research, Special Issue No. 106, pp. 255–258. Coconut Creek (Florida), ISSN 0749-0208.An objective, fair, and accurate evaluation of the financial performance of the port enterprises is an important measure to mobilize and improve the enthusiasm of the port enterprises employees and the progress of scientific and technological innovation. This paper proposes a refined evaluation model of the financial performance of port enterprise companies based on the radial basis function (RBF) neural network. The normalized financial performance index data of port enterprise companies are multiplied by the corresponding weight coefficient as the network input, using excellent, good, middle, and F5 grade evaluations as output. The particle swarm optimization algorithm was used to cross-validate the RBF network structure parameters. Through analyzing the RBF network structure and enter and export characteristics, it was found that the trained RBF network weights are highly correlated with the five-level evaluation results and can more accurately distinguish scientific research performance differences than the five-level evaluation results. This weight can be directly used to conduct a detailed evaluation of the financial performance of the port enterprises. This paper promotes the application of RBF networks in the evaluation of financial performance of port enterprise companies and provides new ideas for similar evaluations.

Early warning of debris flow using optimized self-organizing feature mapping network

Abstract Early warning of debris flow is one of the core contents of disaster prevention and mitigation work for debris flow disasters. There are few early warning methods based on the combination of rainfall threshold and geological environment conditions. In this paper, we presented an early warning method for debris flow based on the infinite irrelevance method (IIM) and self-organizing feature mapping (SOFM), and applied it to Liaoning Province, China. The proposed model consisted of three stages. Firstly, eight geological environmental conditions and two rainfall-inducing conditions were selected by analyzing the factors affecting the development of debris flow in the study area, and the rainfall threshold for debris flow outbreak was 150 mm. Secondly, the correlation between various factors was analyzed by IIM, which prevented the blindness of parameter selection and improved the prediction accuracy of the model. Finally, SOFM was employed to predict the test data. Experimental results showed that the IIM-SOFM model had a strong early warning ability. When 25 samples of low-frequency debris flow area were selected, the accuracy rate of the IIM-SOFM model with optimized network structure parameters was 100%, which it was obviously superior to the rainfall threshold method, BP neural network and competitive neural network. Consequently, it is feasible to use the IIM-SOFM model for early warning of debris flow, outperforming traditional machine learning methods.

Computational fluid dynamic–discrete element method coupling analysis of particle transport in branched networks

An understanding of the particle transport characteristics in a branched network helps to predict the particle distribution and prevent undesired plugging in various engineering systems. Quantitative analysis of particle flow characteristics is challenging in that experiments are expensive and particle flow is difficult to detect without disturbing the flow. To overcome this difficulty, man-made fractal tree-like branched networks were built, and a coupled computational fluid dynamic and discrete element method model was applied. A series of numerical simulations was carried out to analyze the influence of fractal structure parameters of networks on the particle flow characteristics. The joint influence of inertial, shunt capacity and superposition from upstream branches on particle flow was investigated. The injection position at the inlet determined the particle velocity and its future flow path. The particle density ratio, particle size and bifurcation angle had a greater influence on the shunting of K2 branches than that in the K1 level and Nk22/Nk21 reached a maximum at 60°. Compared with a network with an even number of branches, there was a preferential branch when the branch number was odd. The preferential branch effect or asymmetry degree of the level (K2) branches had a more significant impact on particle shunting than that from the upstream branches (K1).

A Forecasting Method of District Heat Load Based on Improved Wavelet Neural Network

Abstract Energy conservation of urban district heating system is an important part of social energy conservation. In response to the situation that the setting of heat load in the system is unreasonable, the heat load forecasting method is adopted to optimize the allocation of resources. At present, the artificial neural networks (ANNs) are generally used to forecast district heat load. In order to solve the problem that networks convergence is slow or even not converged due to the random initial parameters in traditional wavelet neural networks (WNNs), the genetic algorithm with fast convergence ability is used to optimize the network structure and initial parameters of heat load prediction models. The results show that when the improved WNN is applied to forecast district heat load, the prediction error is as low as 2.93%, and the accuracy of prediction results is improved significantly. At the same time, the stability and generalization ability of the prediction model are improved.

Optimal Interplay between Synaptic Strengths and Network Structure Enhances Activity Fluctuations and Information Propagation in Hierarchical Modular Networks.

In network models of spiking neurons, the joint impact of network structure and synaptic parameters on activity propagation is still an open problem. Here, we use an information-theoretical approach to investigate activity propagation in spiking networks with a hierarchical modular topology. We observe that optimized pairwise information propagation emerges due to the increase of either (i) the global synaptic strength parameter or (ii) the number of modules in the network, while the network size remains constant. At the population level, information propagation of activity among adjacent modules is enhanced as the number of modules increases until a maximum value is reached and then decreases, showing that there is an optimal interplay between synaptic strength and modularity for population information flow. This is in contrast to information propagation evaluated among pairs of neurons, which attains maximum value at the maximum values of these two parameter ranges. By examining the network behavior under the increase of synaptic strength and the number of modules, we find that these increases are associated with two different effects: (i) the increase of autocorrelations among individual neurons and (ii) the increase of cross-correlations among pairs of neurons. The second effect is associated with better information propagation in the network. Our results suggest roles that link topological features and synaptic strength levels to the transmission of information in cortical networks.

Modelling the mechanical behaviors of double-network hydrogels

Double-network hydrogels (DN gels), consisting of a rigid, highly cross-linked short-chain network and a flexible, loosely cross-linked long-chain network, are a kind of promising soft and tough material with superior mechanical properties. In order to capture the mechanical behaviors of DN gels more accurately and with more physical depth than some existing models, a constitutive formulation is developed in this paper. The Helmholtz free energy function is decomposed into three parts accounting for the free energies of each network and the mixing of all the components. The eight-chain worm-like chain (WLC) based model is then employed to describe the deformation of each network with two network structure parameters: the chain density and the average chain length. Based on the existing experimental investigations of the internal fracture process of DN gels, a detailed and clear physical picture is illustrated. According to this damage picture, the two network structure parameters of the short-chain network are regarded as two internal variables. Within the framework of general thermodynamics, the energy dissipation is defined through the variation of the internal variables, which allows one to obtain the evolutions of internal variables from analyzing the macroscopic dissipated behaviors of DN gels. For the cases without energy dissipation, the network structure parameters are constant and can be determined directly through the stress-strain relation at one typical point on the stress-strain curve and the stored Helmholtz free energy of each state. These determined values of the parameters are then used to validate the mass conservation. For the cases with energy dissipation, the evolution equations of the two internal variables are determined through the analysis of the energy dissipation and the mass conservation. As a result, not only the unloading stress-stain curves but also the primary loading curve can be theoretically predicted by the constitutive formulation. The mechanical behaviors of both t-DN gels and c-DN gels before necking are analyzed. The comparisons of the theoretical results with the experimental data show sufficiently good agreements, demonstrating the ability and efficiency of the formulation to capture the Mullins effect of DN gels.

Classification Modeling Method for Near-Infrared Spectroscopy of Tobacco Based on Multimodal Convolution Neural Networks.

The origin of tobacco is the most important factor in determining the style characteristics and intrinsic quality of tobacco. There are many applications for the identification of tobacco origin by near-infrared spectroscopy. In order to improve the accuracy of the tobacco origin classification, a near-infrared spectrum (NIRS) identification method based on multimodal convolutional neural networks (CNN) was proposed, taking advantage of the strong feature extraction ability of the CNN. Firstly, the one-dimensional convolutional neural network (1-D CNN) is used to extract and combine the pattern features of one-dimensional NIRS data, and then the extracted features are used for classification. Secondly, the one-dimensional NIRS data are converted into two-dimensional spectral images, and the structure features are extracted from two-dimensional spectral images by the two-dimensional convolutional neural network (2-D CNN) method. The classification is performed by the combination of global and local training features. Finally, the influences of different network structure parameters on model identification performance are studied, and the optimal CNN models are selected and compared. The multimodal NIR-CNN identification models of tobacco origin were established by using NIRS of 5,200 tobacco samples from 10 major tobacco producing provinces in China and 3 foreign countries. The classification accuracy of 1-D CNN and 2-D CNN models was 93.15% and 93.05%, respectively, which was better than the traditional PLS-DA method. The experimental results show that the application of 1-D CNN and 2-D CNN can accurately and reliably distinguish the NIRS data, and it can be developed into a new rapid identification method of tobacco origin, which has an important promotion value.

A OPTIMIZATION TUNED ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM FOR DAM DEFORMATION PREDICTION

Abstract. The performance and stability of Adaptive Neuro-Fuzzy Inference System (ANFIS) depend on its network structure and preset parameter selection, and Particle Swarm Optimization-ANFIS (PSO-ANFIS) easily falls into the local optimum and is imprecise. A novel ANFIS algorithm tuned by Chaotic Particle Swarm Optimization (CPSO-ANFIS) is proposed to solve these problems. A chaotic ergodic algorithm is first used to improve the PSO and obtain a CPSO algorithm, and then the CPSO is used to optimize the parameters of ANFIS to avoid falling into the local optimum and improve the performance of ANFIS. Based on the deformation data from the Xiaolangdi Dam in China, three neural network algorithms, ANFIS, PSO-ANFIS, and CPSO-ANFIS, are used to establish the dam deformation prediction models after data preparation and selection of influencing factors for the dam deformation. The results are compared using evaluation indicators that show that CPSO-ANFIS is more accurate and stable than ANFIS and PSO-ANFIS both in predictive ability and in predicted results.

A data-driven operational integrated driving behavioral model on highways

Car-following (CF) and lane-changing (LC) behavior models have been widely studied separately as the core models of traffic simulation. However, in practice, CF and LC are inseparable and thus integrated driving (ID) models containing CF and LC behaviors emerge. Here, we proposed a new work to introduce the social force (SF) model to the operational ID behavioral model on the highway. First, a data-driven-based operational ID behavioral model is proposed in the hierarchical social force behavioral model framework. Then, the inputs/output of the SF-ID behavioral model is determined. SF-ID model is built by the feed-forward neural networks (FNN), and the network structure and other parameters are calibrated and verified by field data. Results of the test on CF and LC situations show that our proposed FNN SF-ID model has a good capability in reproducing/predicting the operational ID behaviors on the highway. In addition, we also analyzed the structural features of the FNN SF-ID models, and refine the original models by removing insignificant inputs. The comparison results showed that the refined model—FNN SF-ID (R)—performed better than the original models.

Assessment of pre and post-disaster supply chain resilience based on network structural parameters with CVaR as a risk measure

The present study assesses supply chain resilience based on network structural parameters. Resilience is computed as a composite effect of density, centrality, connectivity, and network size of the network. A simulation-based approach is adopted, wherein networks of 23 firms operating in India are subjected to risk combinations of five mutually inclusive independent scenarios of probability levels and five mutually exclusive and exhaustive impact levels. The worst-case performance of the supply chain network when subjected to high impact and low probability risks is captured using conditional-value at risk (CVaR). Results reveal that the firm which has the lowest density and centrality and the highest connectivity and network size, exhibits the highest resilience. Whereas, the firm which has the highest density and high centrality due to an aggregation node exhibits the lowest resilience.The two main contributions of the present study are as follows. First, it derives insights for practicing managers from actual instead of theoretical networks. Second, it captures the worst-case performance of the supply chain network using CVaR, which has not been reported by any study in the supply chain network structure domain. The simulation-based approach can be easily adopted by the managers to assess the resilience of their supply chain networks and their preparedness to face potential risks. The information available in the form of CVaR is an important input to practicing managers to evaluate whether their supply chain network can face severe disruptions or not. The managers can then make informed decisions on how to increase the resilience of their supply chain networks.

Mixed-integer minimax dynamic optimization for structure identification of glycerol metabolic network

Cell metabolism is a dynamic regulation process, in which its network structure and/or regulatory mechanisms can change constantly over time due to internal and external perturbations. This paper models glycerol metabolism in continuous fermentation as a nonlinear mixed-integer dynamic system by defining the time-varying metabolic network structure as an integer-valued function. To identify the dynamic network structure and kinetic parameters, we establish a mixed-integer minimax dynamic optimization problem with concentration robustness as its objective functional. By direct multiple shooting strategy and a decomposition approach consisting of convexification, relaxation and rounding strategy, the optimization problem is transformed into a large-scale approximate multistage parameter optimization problem. It is then solved using a competitive particle swarm optimization algorithm. We also show that the relaxation problem yields the best lower bound for the optimization problem, and its solution can be arbitrarily approximated by the solution obtained from rounding strategy. Numerical results indicate that the proposed mixed-integer dynamic system can better describe cellular self-regulation and response to intermediate metabolite inhibitions in continuous fermentation of glycerol. These numerical results show that the proposed numerical methods are effective in solving the large-scale mixed-integer dynamic optimization problems.

Social network analysis in pigs: impacts of significant dyads on general network and centrality parameters

In general, one animal is considered dominant over another animal if it has won more fights than its opponent. Whether this difference in won and lost fights is significant is neglected in most studies. Thus, the present study evaluates the impact of two different calculation methods for dyadic interactions with a significant asymmetric outcome on the results of social network analysis regarding agonistic interactions of pigs in three different mixing events (weaned piglets, fattening pigs and gilts). Directly after mixing, all animals were video recorded for 17 (fattening pigs, gilts) and 28 h (weaned piglets), documenting agonistic interactions. Two calculation methods for significant dyads, that is, dyadic interactions with a clear dominant subordinate relationship in which one animal has won significantly more fights than its encounter, were proposed: pen individual limits were calculated by a sign test considering the differences of won and lost fights of all dyadic interactions in each pen; dyad individual limits were determined by a one-sided sign test for each individual dyad. For all data sets (ALL, including all dyadic interactions; PEN or DYAD, including only significant dyads according to pen or dyad individual limits), networks were built based on the information of initiator and receiver with the pigs as nodes and the edges between them illustrating attacks. General network parameters describing the whole network structure and centrality parameters describing the position of each animal in the network were calculated. Both pen and dyad individual limits revealed only a small percentage of significant dyads for weaned piglets (12.4% or 8.8%), fattening pigs (4.2% or 0.6%) and gilts (3.6% or 0.4%). The comparison between the data sets revealed only high Spearman’s rank correlation coefficients (rS) for the density, that is, percentage of possible edges that were actually present in the network, whereas the centrality parameters showed only moderate rS values (0.37 to 0.75). Thus, the rank order of the animals changed due to the exclusion of insignificant dyads, which shows that the results obtained from social network analysis are clearly influenced if insignificant dyads are excluded from the analyses. Due to the fact that the pen individual limits consider the overall level of agonistic interactions within each pen, this calculation method should be preferred over the dyad individual limits. Otherwise, too many animals in the group became isolated nodes with zero centrality for which no statement about their position within the network can be made.

The Symptom and Domain Structure of Psychotic Disorders: A Network Analysis Approach

Abstract Little is understood about the symptom network structure of psychotic disorders. In the current study, we aimed to examine the network structure of psychotic symptoms in a broad and transdiagnostic sample of subjects with psychotic disorders (n = 2240) and to determine whether network structure parameters vary across demographic, sampling method and clinical variables. Gaussian graphical models were estimated for 73 psychotic symptoms assessed using the Comprehensive Assessment of Symptoms and History. A 7-cluster solution (reality distortion, disorganization, catatonia, diminished expressivity, avolition/anhedonia, mania, and depression) best explained the underlying symptom structure of the network. Symptoms with the highest centrality estimates pertained to the disorganization and, to a lesser extent, negative domains. Most bridge symptoms pertained to the disorganization domain, which had a central position within the network and widespread connections with other psychopathological domains. A comparison of networks in subgroups of subjects defined by premorbid adjustment levels, treatment response, and course pattern significantly influenced both network global strength and network structure. The sampling method and diagnostic class influenced network structure but not network global strength. Subgroups of subjects with less densely connected networks had poorer outcomes or more illness severity than those with more densely connected networks. The network structure of psychotic features emphasizes the importance of disorganization symptoms as a central domain of psychopathology and raises the possibility that interventions that target these symptoms may prove of broad use across psychopathology. The network structure of psychotic disorders is dependent on the sampling method and important clinical variables.

Statistical Process Control with Intelligence Based on the Deep Learning Model

Statistical process control (SPC) is an important tool of enterprise quality management. It can scientifically distinguish the abnormal fluctuations of product quality. Therefore, intelligent and efficient SPC is of great significance to the manufacturing industry, especially in the context of industry 4.0. The intelligence of SPC is embodied in the realization of histogram pattern recognition (HPR) and control chart pattern recognition (CCPR). In view of the lack of HPR research and the complexity and low efficiency of the manual feature of control chart pattern, an intelligent SPC method based on feature learning is proposed. This method uses multilayer bidirectional long short-term memory network (Bi-LSTM) to learn the best features from the raw data, and it is universal to HPR and CCPR. Firstly, the training and test data sets are generated by Monte Carlo simulation algorithm. There are seven histogram patterns (HPs) and nine control chart patterns (CCPs). Then, the network structure parameters and training parameters are optimized to obtain the best training effect. Finally, the proposed method is compared with traditional methods and other deep learning methods. The results show that the quality of extracted features by multilayer Bi-LSTM is the highest. It has obvious advantages over other methods in recognition accuracy, despite the HPR or CCPR. In addition, the abnormal patterns of data in actual production can be effectively identified.

Hyperspectral Image Classification Based on 3-D Separable ResNet and Transfer Learning

Deep learning (DL) has proven to be a promising technique for hyperspectral image (HSI) classification. However, due to complex network structure and massive parameters, it is challenging to achieve satisfying classification accuracy with only a small number of training samples. In this letter, we propose a novel HSI classification method by collaborating the 3-D separable ResNet (3-D-SRNet) with cross-sensor transfer learning. The 3-D-SRNet replaces 3-D convolutions with spatial and spectral separable 3-D convolutions, thus showing much less parameters than models that use standard 3-D convolutions. First, we pretrain a classification model with the proposed 3-D-SRNet on the source HSI data set with sufficient training samples compared with the target HSI data set. Then, the pretrained model is transferred to the target HSI data set for fine-tuning to finish the classification task. It is worth noting that the source data for pretraining can be captured by the different sensor with the target data. Compared with the conventional 3-D-ResNet, the proposed 3-D-SRNet has less parameters involving lower computation cost while achieving better classification performance. Experimental results on three benchmark data sets show that our method outperforms several state-of-the-art methods in HSI classification with small training samples.

NANUQ: a method for inferring species networks from gene trees under the coalescent model

Species networks generalize the notion of species trees to allow for hybridization or other lateral gene transfer. Under the network multispecies coalescent model, individual gene trees arising from a network can have any topology, but arise with frequencies dependent on the network structure and numerical parameters. We propose a new algorithm for statistical inference of a level-1 species network under this model, from data consisting of gene tree topologies, and provide the theoretical justification for it. The algorithm is based on an analysis of quartets displayed on gene trees, combining several statistical hypothesis tests with combinatorial ideas such as a quartet-based intertaxon distance appropriate to networks, the NeighborNet algorithm for circular split systems, and the Circular Network algorithm for constructing a splits graph.

Facial Expression Recognition Method Based on Improved VGG Convolutional Neural Network

Because the shallow neural network has limited ability to represent complex functions with limited samples and calculation units, its generalization ability will be limited when it comes to complex classification problems. The essence of deep learning is to learn a nonlinear network structure, to represent input data distributed representation and demonstrate a powerful ability to learn deeper features of data from a small set of samples. In order to realize the accurate classification of expression images under normal conditions, this paper proposes an expression recognition model of improved Visual Geometry Group (VGG) deep convolutional neural network (CNN). Based on the VGG-19, the model optimizes network structure and network parameters. Most expression databases are unable to train the entire network from the start due to lack of sufficient data. This paper uses migration learning techniques to overcome the shortage of image training samples. Shallow CNN, Alex-Net and improved VGG-19 deep CNN are used to train and analyze the facial expression data on the Extended Cohn–Kanade expression database, and compare the experimental results obtained. The experimental results indicate that the improved VGG-19 network model can achieve 96% accuracy in facial expression recognition, which is obviously superior to the results of other network models.

Taming chimeras in networks through multiplexing delays

Chimera referring to a coexistence of coherent and incoherent states, is traditionally very difficult to control due to its peculiar nature. Here, we provide a recipe to construct chimera states in the multiplex networks with the aid of multiplexing delays. The chimera state in multiplex networks is produced by introducing heterogeneous delays in a fraction of inter-layer links, referred to as multiplexing delay, in a sequence. Additionally, the emergence of the incoherence in the chimera state can be regulated by making the appropriate choice of both inter- and intra-layer coupling strengths, whereas the extent and the position of the incoherence regime can be regulated by appropriate placing and strength of the multiplexing delays. The proposed technique to construct such engineered chimera equips us with multiplex network's structural parameters as tools in gaining both qualitative and quantitative control over the incoherent section of the chimera states and, in turn, the chimera. Our investigation can be of worth in controlling dynamics of multi-level delayed systems and attain the desired chimeric patterns.

Classifying transportation mode and speed from trajectory data via deep multi-scale learning

With the rapid development of mobile Internet, the Internet of Things and other new technologies, mobile devices are generating massive amounts of spatio-temporal trajectory data. This paper aims to propose a method that can automatically classify transportation mode and speed, help people understand the mobility of moving objects, thus making people’s life more convenient and traffic management easier. Although there have been some studies on trajectory classification, yet they either require manual feature selection or fail to fully consider the impact of time and space on classification results. None of them can extract features automatically and comprehensively. Hence, we propose Deep Multi-Scale Learning Model and design a deep neural network to learn features under multi-scale time and space granularities automatically. The obtained features are fused to output final classification results. Our method is based on the latest image classification network structure DenseNet, and incorporates attention mechanism and residual learning. This model is able to fully capture spatial features so as to enhance feature propagation and capture long-term dependence. Moreover, the number of network structure parameters is also reduced. We have evaluated our Deep Multi-Scale Learning Model on two real datasets. The results show that our model is superior to the current state-of-the-art models in top-1 accuracy, recall and f1-score. Furthermore, the classification results from our model can help to understand mobility accurately.

A new genetically optimized tensor product functional link neural network: an application to the daily exchange rate forecasting

The training speed for multilayer neural networks is slow due to the multilayering. Therefore, removing the hidden layers, provided that the input layer is endowed with additional higher order units is suggested to avoid such problem. Tensor product functional link neural network (TPFLNN) is a single layer with higher order terms that extend the network’s structure by introducing supplementary inputs to the network (i.e., joint activations). Although the structure of the TPFLNN is simple, it suffers from weight combinatorial explosion problem when its order becomes excessively high. Furthermore, similarly to many neural network methods, selection of proper weights is one of the most challenging issues in the TPFLNN. Finding suitable weights could help to reduce the number of needed weights. Therefore, in this study, the genetic algorithm (GA) was used to find near-optimum weights for the TPFLNN. The proposed method is abbreviated as GA–TPFLNN. The GA–TPFLNN was used to forecast the daily exchange rate for the Euro/US Dollar, and Japanese Yen/US Dollar. Simulation results showed that the GA–TPFLNN produced more accurate forecasts as compared to the standard TPFLNN, GA, GA–TPFLNN with backpropagation, GA-functional expansion FLNN, multilayer perceptron, support vector regression, random forests for regression, and naive methods. The GA helps the TPFLNN to find low complexity network structure and/or near-optimum parameters which leads to this better result.