交通事故的时空热点分析

针对目前交通事故分析中时空维度分离的不足，以H市2013―2015年的交通事故数据为研究对象，根据自组织神经网络、平行坐标系、时空颜色矩阵和时空网络核密度估计、热点分析法，分别从行政区划以及道路网络上进行交通事故的时空特征分析，从宏观和微观角度揭示交通事故的时空热点区域。结果表明，整体来说，在空间上，H市交通事故热点区域4个街道高于8个镇，呈明显的“两带一中心”分布，即硖许线与101省道形成的带状区域和市政府所在的行政中心区域；时间上，早、晚高峰时段最为严重，夜晚较为轻微，呈明显的区域特征分布，即工业办公区热点出现在早高峰时段，生活住宅区热点出现在晚高峰时段，商业消费区热点出现在夜晚时段。     

Pinning synchronization with low energy cost

The pinning synchronization on complex networks with low energy cost was investigated. Two basic problems were considered, in this paper. One is how to achieve synchronization on complex network with weak coupling. Another is how to reduce the energy cost in the process of achieving and keeping synchronization. On the other hand, for a good pinning strategy, the synchronizing speed should also be ensured. In this paper, a switching pinning strategy was proposed to enhance the synchronizing speed. And from the point of view of energy saving, some selected links were strengthened. Theoretical analysis and simulations on different complex networks indicate that the proposed pinning strategy is effective. Compare with conventional existing method, the superiority of our method is significant.     

新组合模型在光电功率预测中的应用

为提高光伏预测要求的精准性,文章提出一种新算法将神经网络和ARMA算法改进组合,构成NEW ARMA-BP模型算法.以某30兆瓦的光伏电站采集的输出功率为输入样本,基于ARMA和BP神经网络算法在Matlab环境下依次搭建了相应的预测模型,预估光伏短期输出量.采用"误差正态检验图"判断基于两种不同算法的误差水平,依据两种单模型预测误差,运用所提出的新方法计算权值并获得新的预测值.基于Matlab的仿真结论验证了组合预测在光伏输出预测领域的适用性.     

在线社交网络控制实验的现状与展望

在线社交网络已发展成为一个独特的电子生态系统，其应用深刻影响着人们生活的方方面面。由于在线社交网络特性复杂，分析在线社交网络形成和变化中的规律成为当前计算机科学、社会学和物理学的一项挑战。传统上，在线社交网络实证研究主要采用计算机辅助的被动数据获取和分析方式。近年来，在真实大规模在线社交网络上直接进行控制实验从而主动获取数据并开展分析研究的方式广受关注。评述了这一领域的研究进展，包括：社交网络控制实验的主要研究模式；控制实验方法在社交网络结构、信息传播、行为和心理学等领域取得的主要成果以及主要实验工具的适用条件和局限性。最后，展望了人工智能技术在社交网络控制实验中的应用潜力，分析了智能算法对降低实验成本和提高实验效率的作用。     

Self‐Assembly of a Tripod Aromatic Rod into Stacked Planar Networks

Threefold symmetric rigid‐core molecules with an internally grafted poly(ethylene oxide) (PEO) chain were synthesized, and their self‐assembled structures were characterized using differential scanning calorimetry, TEM, and 1D and 2D X‐ray scatterings in the solid state. The tripod compounds based on short PEO chains (n=8, 13, 17, 21), self‐assemble into 2D channel‐like network structures, whereas the compound with the longest PEO chain (n=34) forms a lamellar liquid crystalline phase. The interiors of the channel structures are filled with flexible PEO chains along the double‐walled aromatic circumference. In these channel‐like networks, three aromatic rods connected in the meta‐position to each other are superimposed in parallel to other adjacent molecules to form the double‐walled aromatic frameworks stacked perpendicular to the resulting channels. These are novel examples of supramolecular channel‐like structures developed using amphiphilic diblock molecules based on a threefold symmetric rigid scaffold.     

Prediction of Concrete Meso-Model Stress-Strain Curves Based on GoogLeNet北大核心CSCD

非均质复合材料的宏观力学性能往往取决于细观组分的分布方式和力学性能，但是建立明确的关系表达式极其困难。为了应对这一挑战，以混凝土为研究对象，提出了一种基于深度学习的策略，能够高效、准确地通过细观模型图像信息获取应力-应变曲线。首先，使用基于卷积神经网络（convolutional neural network，CNN）的GoogLeNet模型进行图像信息识别和提取，并针对应力-应变曲线的复杂性特点，进行了数据预处理操作，并且设计了相应的多任务损失函数。数据集中的细观模型图像采用基于Monte-Carlo的随机骨料模型生成，并且使用数值模拟试验获取对应细观模型的单轴压缩应力-应变曲线。最后，通过对神经网络的训练和测试评估了所提出方法的可行性。结果表明，GoogLeNet模型训练效率和预测精度均优于AlexNet和ResNet模型，具有良好的泛化能力和鲁棒性。     

团体随机和择优混合演化网络模型的稳定性

主要通过马氏链、主方程的方法和技巧,给出了团体随机和择优混和演化网络的稳态度分布存在性的严格证明,并严格推导了度分布的精确解析表达式.     

From curing kinetics to network structure: A novel approach to the modeling of the network buildup of epoxy–anhydride thermosets

The curing kinetics and network buildup during curing of epoxy–anhydride formulations using tertiary amines as initiators are reviewed in this work. A mechanism‐based kinetic and structural model has been defined, showing better prediction capabilities than previous living polymerization and simple regeneration models. The curing kinetics have been analyzed using differential scanning calorimetry (DSC), and the gelation during curing has been determined by combined thermomechanical analysis and DSC. The effect of initiator content and epoxy equivalent weight are taken into account. The network buildup has been modeled using a stochastic network buildup model based on the random combination of primary chains generated by the kinetic model. © 2013 Wiley Periodicals, Inc. J. Polym. Sci., Part A: Polym. Chem. 2014 , 52, 61–75     

Viscoelastic properties of supramolecular soft materials with transient polymer network

A series of supramolecular soft materials with hydrogen bonded transient networks was prepared by blending carboxy‐terminated telechelic poly(ethyl acrylate) (PEA‐(COOH)₂) and polyethyleneimine (PEI). Effects of PEA‐(COOH)₂ molecular weight (M_PEA) and the blend ratio on the viscoelastic properties were investigated by rheological and small angle X‐ray scattering measurements. Rubbery plateau appeared by adding PEI due to network formation with ionic hydrogen bonded crosslinks between amines on PEI and carboxylic acids on PEA‐(COOH)₂. The highest temperature of a storage modulus‐loss modulus crossover as well as the highest flow activation energy was attained at a certain mole ratio of amines to carboxylic acids, irrelevant to M_PEA, indicating optimized supramolecular networks were achieved by stoichiometric balance of two functional groups. Since telechelic PEA‐(COOH)₂ serves as a network strand, the plateau modulus was inversely proportional to M_PEA, which was consistent with the correlation length between crosslinks estimated by X‐ray scattering measurements. © 2014 Wiley Periodicals, Inc. J. Polym. Sci., Part B: Polym. Phys. 2014 , 52, 755–764     

Analytical solution of stochastic differential equation by multilayer perceptron neural network approximation of Fokker–Planck equation

The Fokker–Planck equation is a useful tool to analyze the transient probability density function of the states of a stochastic differential equation. In this paper, a multilayer perceptron neural network is utilized to approximate the solution of the Fokker–Planck equation. To use unconstrained optimization in neural network training, a special form of the trial solution is considered to satisfy the initial and boundary conditions. The weights of the neural network are calculated by Levenberg–Marquardt training algorithm with Bayesian regularization. Three practical examples demonstrate the efficiency of the proposed method.