基于多粒度结构的网络表示学习

图卷积网络(GCN)能够适应不同结构的图,但多数基于GCN的方法难以有效地捕获网络的高阶相似性。简单添加卷积层将导致输出特征过度平滑并使它们难以区分,而且深层神经网络更难训练。本文选择将网络的多粒度结构和图卷积网络结合起来用于学习网络的节点特征表示,提出基于多粒度结构的网络表示学习方法Multi-GS。首先,基于模块度聚类和粒计算思想,用分层递阶的多粒度空间替代原始的单层网络拓扑空间;然后,利用GCN模型学习不同粗细粒度空间中粒的表示;最后,由粗到细将不同粒的表示组合为原始空间中节点的表示。实验结果表明:Multi-GS能够捕获多种结构信息,包括一阶和二阶相似性、社团内相似性(高阶结构)和社团间相似性(全局结构)。在绝大多数情况下,使用多粒度的结构可改善节点分类任务的分类效果。

Innovative approaches in high-speed railway bridge model simplification for enhanced computational efficiency

In the realm of high-speed railway bridge engineering,managing the intricacies of the track-bridge system model(TBSM)during seismic events remains a formidable challenge.This study pioneers an innovative approach by presenting a simplified bridge model(SBM)optimized for both computational efficiency and precise representation,a seminal contribution to the engineering design landscape.Central to this innovation is a novel model-updating methodology that synergistically melds artificial neural networks with an augmented particle swarm optimization.The neural networks adeptly map update parameters to seismic responses,while enhancements to the particle swarm algorithm’s inertial and learning weights lead to superior SBM parameter updates.Verification via a 4-span high-speed railway bridge revealed that the optimized SBM and TBSM exhibit a highly consistent structural natural period and seismic response,with errors controlled within 7%.Additionally,the computational efficiency improved by over 100%.Leveraging the peak displacement and shear force residuals from the seismic TBSM and SBM as optimization objectives,SBM parameters are adeptly revised.Furthermore,the incorporation of elastoplastic springs at the beam ends of the simplified model effectively captures the additional mass,stiffness,and constraint effects exerted by the track system on the bridge structure.

Springback prediction for incremental sheet forming based on FEM-PSONN technology

In the incremental sheet forming （ISF） process, springback is a very important factor that affects the quality of parts. Predicting and controlling springback accurately is essential for the design of the toolpath for ISF. A three-dimensional elasto-plastic finite element model （FEM） was developed to simulate the process and the simulated results were compared with those from the experiment. The springback angle was found to be in accordance with the experimental result, proving the FEM to be effective. A coupled artificial neural networks （ANN） and finite element method technique was developed to simulate and predict springback responses to changes in the processing parameters. A particle swarm optimization （PSO） algorithm was used to optimize the weights and thresholds of the neural network model. The neural network was trained using available FEM simulation data. The results showed that a more accurate prediction of s!oringback can be acquired using the FEM-PSONN model.

Modified particle swarm optimization-based antenna tilt angle adjusting scheme for LTE coverage optimization

In order to solve the challenging coverage problem that the long term evolution（ LTE） networks are facing, a coverage optimization scheme by adjusting the antenna tilt angle（ ATA） of evolved Node B（ e NB） is proposed based on the modified particle swarm optimization（ MPSO） algorithm.The number of mobile stations（ MSs） served by e NBs, which is obtained based on the reference signal received power（RSRP） measured from the MS, is used as the metric for coverage optimization, and the coverage problem is optimized by maximizing the number of served MSs. In the MPSO algorithm, a swarm of particles known as the set of ATAs is available; the fitness function is defined as the total number of the served MSs; and the evolution velocity corresponds to the ATAs adjustment scale for each iteration cycle. Simulation results showthat compared with the fixed ATA, the number of served MSs by e NBs is significantly increased by 7. 2%, the quality of the received signal is considerably improved by 20 d Bm, and, particularly, the system throughput is also effectively increased by 55 Mbit / s.

A novel internet traffic identification approach using wavelet packet decomposition and neural network

Internet traffic classification plays an important role in network management, and many approaches have been proposed to classify different kinds of internet traffics. A novel approach was proposed to classify network applications by optimized back-propagation (BP) neural network. Particle swarm optimization (PSO) algorithm was used to optimize the BP neural network. And in order to increase the identification performance, wavelet packet decomposition (WPD) was used to extract several hidden features from the time-frequency information of network traffic. The experimental results show that the average classification accuracy of various network applications can reach 97%. Moreover, this approach optimized by BP neural network takes 50% of the training time compared with the traditional neural network.

Prediction of Flash Point Temperature of Organic Compounds Using a Hybrid Method of Group Contribution ＋ Neural Network ＋ Particle Swarm Optimization

The flash points of organic compounds were estimated using a hybrid method that includes a simple group contribution method (GCM) implemented in an artificial neural network (ANN) with particle swarm optimization (PSO). Different topologies of a multilayer neural network were studied and the optimum architecture was determined. Property data of 350 compounds were used for training the network. To discriminate different substances the molecular structures defined by the concept of the classical group contribution method were given as input variables. The capabilities of the network were tested with 155 substances not considered in the training step. The study shows that the proposed GCM+ANN+PSO method represent an excellent alternative for the estimation of flash points of organic compounds with acceptable accuracy (AARD = 1.8%; AAE = 6.2 K).

Boiler combustion optimization based on ANN and PSO-Powell algorithm

To improve the thermal efficiency and reduce nitrogen oxides （NOx ） emissions in a power plant for energy conservation and environment protection, based on the reconstructed section temperature field and other related parameters, dynamic radial basis function （RBF） artificial neural network （ANN） models for forecasting unburned carbon in fly ash and NO, emissions in flue gas ware developed in this paper, together with a multi-objective optimization system utilizing particle swarm optimization and Powell （PSO-Powell） algorithm. To validate the proposed approach, a series of field tests were conducted in a 350 MW power plant. The results indicate that PSO-Powell algorithm can improve the capability to search optimization solution of PSO algorithm, and the effectiveness of system. Its prospective application in the optimization of a pulverized coal （ PC ） fired boiler is presented as well.

A fuzzy neural network evolved by particle swarm optimization

A cooperative system of a fuzzy logic model and a fuzzy neural network(CSFLMFNN)is proposed,in which a fuzzy logic model is acquired from domain experts and a fuzzy neural network is generated and prewired according to the model.Then PSO-CSFLMFNN is constructed by introducing particle swarm optimization(PSO)into the cooperative system instead of the commonly used evolutionary algorithms to evolve the prewired fuzzy neural network.The evolutionary fuzzy neural network implements accuracy fuzzy inference without rule matching.PSO-CSFLMFNN is applied to the intelligent fault diagnosis for a petrochemical engineering equipment,in which the cooperative system is proved to be effective.It is shown by the applied results that the performance of the evolutionary fuzzy neural network outperforms remarkably that of the one evolved by genetic algorithm in the convergence rate and the generalization precision.

Catalytic Cracking and PSO-RBF Neural Network Model of FCC Cycle Oil

Catalytic cracking experiments of FCC cycle oil were carried out in a fixed fluidized bed reactor. Effects of reac- tion conditions, such as temperature, catalyst to oil ratio and weight hourly space velocity, were investigated. Hydrocarbon composition of gasoline was analyzed by gas chromatograph. Experimental results showed that conversion of cycle oil was low on account of its poor crackability performance, and the effect of reaction conditions on gasoline yield was obvi- ous. The paraffin content was very high in gasoline. Based on the experimental yields under different reaction conditions, a model for prediction of gasoline and diesel yields was established by radial basis function neural network （RBFNN）. In the model, the product yield was viewed as function of reaction conditions. Particle swarm optimization （PSO） algorithm with global search capability was used to obtain optimal conditions for a highest yield of light oil. The results showed that the yield of gasoline and diesel predicted by RBF neural network agreed well with the experimental values. The optimized reac- tion conditions were obtained at a reaction temperature of around 520 ~C, a catalyst to oil ratio of 7.4 and a space velocity of 8 h~. The predicted total yield of gasoline and diesel reached 42.2% under optimized conditions.

ESTIMATION OF ROCK-AGGREGATE VOLUME BASED ON PCA AND LM-OPTIMIZED NEURAL NETWORK

In granule processing industries, acquisition of particle size and shape parameters is a common procedure, and volumetric measurement is of great importance in dealing with particle sizing and gradation. To eradicate the major drawbacks with manual gauge, this paper proposes an optical approach using Back Propagation （BP） neural network to estimate the particle volume based on the two-Dimensional （2D） image information. To achieve the better network efficiency and structure simplicity, Principal Component Analysis （PCA） is adopted to reduce the dimensions of network inputs To overcome the shortcomings of generic BP network for being slow to converge and vulnerable to being trapped in local minimum, Levenberg-Marquardt （LM） algorithm is applied to achieve a higher speed and a lower error rate. The real particle data is utilized in training and testing the presented network. The experimental result suggests that the proposed neural network is capable of estimating aggregate volume with satisfactory precision and superior to the generic BP network in terms of perforxnance capacity.

A Fuzzy Neural Network Model of Linguistic Dynamic Systems Based on Computing with Words

Linguistic dynamic systems(LDS)are dynamic processes involving computing with words(CW)for modeling and analysis of complex systems.In this paper,a fuzzy neural network(FNN)structure of LDS was proposed.In addition,an improved nonlinear particle swarm optimization was employed for training FNN.The experiment results on logistics formulation demonstrates the feasibility and the efficiency of this FNN model.

Control of Neural Network Feedback Linearization Based on Chaotic Particle Swarm Optimization

A new chaotic particle swarm algorithm is proposed in order to avoid the premature convergence of the particle swarm optimization and the shortcomings of the chaotic optimization, such as slow searching speed and low accuracy when used in the multivariable systems or in large search space. The new algorithm combines the particle swarm algorithm and the chaotic optimization, using randomness and ergodicity of chaos to overcome the premature convergence of the particle swarm optimization. At the same time, a new neural network feedback linearization control system is built to control the single-machine infinite-bus system. The network parameters are trained by the chaos particle swarm algorithm, which makes the control achieve optimization and the control law of prime mover output torque obtained. Finally, numerical simulation and practical application validate the effectiveness of the method.

MULTI-GRANULARITY EVOLUTION ANALYSIS OF SOFTWARE USING COMPLEX NETWORK THEORY

Software systems are a typical kind of man-made complex systems. Understanding their evolutions can lead to better software engineering practices. In this paper, the authors use complex network theory as a tool to analyze the evolution of object-oriented （OO） software from a multi-granularity perspective. First, a multi-granularity software networks model is proposed to represent the topological structures of a multi-version software system from three levels of granularity. Then, some parameters widely used in complex network theory are applied to characterize the software networks. By tracing the parameters＇ values in consecutive software systems, we have a better understanding about software evolution. A case study is conducted on an open source OO project, Azureus, as an example to illustrate our approach, and some underlying evolution characteristics are uncovered. These results provide a different dimension to our understanding of software evolutions and also are very useful for the design and development of OO software systems.

Bearing capacity of thin-walled shallow foundations： an experimental and artificial intelligence-based study

Thin-walled spread foundations are used in coastal projects where the soil strength is relatively low. Developing a predictive model of bearing capacity for this kind of foundation is of interest due to the fact that the famous bearing capacity equations are proposed for conventional footings. Many studies underlined the applicability of artificial neural networks （ANNs） in predicting the bearing capacity of foundations. However, the majority of these models are built using conventional ANNs, which suffer from slow rate of learning as well as getting trapped in local minima. Moreover, they are mainly developed for conventional footings. The prime objective of this study is to propose an improved ANN-based predictive model of bearing capacity for thin-walled shallow foundations. In this regard, a relatively large dataset comprising 145 recorded cases of related footing load tests was compiled from the literature. The dataset includes bearing capacity （Qu）, friction angle, unit weight of sand, footing width, and thin-wall length to footing width ratio （Lw/B）. Apart from Qu, other parameters were set as model inputs. To enhance the diversity of the data, four more related laboratory footing load tests were conducted on the Johor Bahru sand, and results were added to the dataset. Experimental findings suggest an almost 0.5 times increase in the bearing capacity in loose and dense sands when LJB is increased from 0.5 to 1.12. Overall, findings show the feasibility of the ANN-based predictive model improved with particle swarm optimization （PSO）. The correlation coefficient was 0.98 for testing data, suggesting that the model serves as a reliable tool in predicting the bearing capacity.

Improved Algorithm for Distributed Localization in Wireless Sensor Networks

Many sensor network applications require location awareness,but it is often too expensive to equip a global positioning system(GPS) receiver for each network node.Hence,localization schemes for sensor networks typically use a small number of seed nodes that know their locations and protocols whereby other nodes estimate their locations from the messages they receive.For the inherent shortcomings of general particle filter(the sequential Monte Carlo method) this paper introduces particle swarm optimization and weighted centroid algorithm to optimize it.Based on improvement a distributed localization algorithm named WC-IPF(weighted centroid algorithm improved particle filter) has been proposed for localization.In this localization scheme the initial estimate position can be acquired by weighted centroid algorithm.Then the accurate position can be gotten via improved particle filter recursively.The extend simulation results show that the proposed algorithm is efficient for most condition.