基于WFC和MI的主题句提取方法

提出一种基于加权模糊聚类（WFC）和互信息（MI）的主题句提取方法,使主题句尽可能全面覆盖全文主题的同时,缩减自身的冗余,以提高摘要效率,采用加权模糊聚类的方法对文本句子进行分类,对在同一类中的句子使用比较互信息的方法进行排名处理,从而获得高质量的摘要。实验结果表明,与传统聚类方法比较,该方法的正确率提高约15%,可以达到约70%的精确度,并在阅读摘要时能够基本正确地获取文本信息。     

基于改进时空残差卷积神经网络的城市路网短时交通流预测

城市路网交通流预测受到历史交通流和相邻路口交通流的影响,具有复杂的时空关联性.针对传统时空残差模型缺乏对交通流数据进行相关性分析、捕获微小变化而容易忽略长期时间特征等问题,提出一种基于改进时空残差卷积神经网络(CNN)的城市路网短时交通流预测模型.该模型将原始交通流数据转化成交通栅格数据,利用皮尔逊相关系数(PCC)对...     

融合子块梯度与线性预测的单高斯背景建模

针对单高斯背景模型不能适应非平稳场景且对初期保持静止后期运动的物体造成"鬼影"现象的问题,提出了融合子块梯度与线性预测的单高斯背景建模方法。首先,对每个像素点进行单高斯背景建模,并实现像素级的自适应更新,运用子块梯度算法将梯度在阈值内的子块作为背景以消除"鬼影";然后,将子块梯度法获得的前景与单高斯模型确定的前景做与运算,提高在非平稳场景下对背景的判断能力;最后,运用线性预测方法处理获得的前景点,将面积小于阈值的连通区域还原为背景。采用CDNET 2012 Dataset和Wallflower Dataset进行仿真实验:当场景变化幅度较大时,所提算法与混合高斯模型(GMM)相比,虽然检测率稍有下降,但检测精度提高了40%;在其他场景中检测率虽只提高约10%,检测精度却能提高25%以上。实验结果表明,融合子块梯度与线性预测的单高斯背景建模能够适应非平稳场景并消除"鬼影"现象,获得的背景比混合高斯模型更精确,提取的前景细节更丰富。     

Developing a design support system for the exterior form of running shoes using partial least squares and neural networks

The function of sport shoes is to improve sport performance and reduce sports-related injuries. They are commercial products developed by combining sports technology and marketing activities. Numerous studies on research and development, material application, production process improvements, and human physiological measurements for the sport shoe market exist. However, few studies have conducted an in-depth investigation on the design of forms and external appearance for sports shoes.     

A 6:1 image‐compression method using directional prediction for LCD overdrive

Abstract— An overdrive technology was developed and is widely used to diminish motion blur in LCDs. To store a previous frame in the overdrive operation in a limited‐sized memory, simple image‐compression techniques are required. By considering the strong correlation of nearby pixels in natural images, a new 6:1 color‐image‐compression method based on directional prediction is proposed. Different from the directional prediction of intra‐coding in H.264/AVC, the predictable direction is determined beforehand to minimize the computation complexity. A simple content‐adaptive quantization and bit‐streaming method, which preserves image details and is free from blocking artifacts, is also proposed. The experimental results show that our proposed method outperforms the vector quantization block truncation coding method with an average 3‐dB peak signal‐to‐noise ratio (PSNR) as well as the subjective quality in terms of blocking artifacts.     

A prediction and auto-execution system of smartphone application services based on user context-awareness

Smartphones have emerged as suitable environments for user context-awareness and intelligent service provision due to the high penetration rate, the high usability, various embedded sensors, and so on. In particular, its most unique characteristic is the usage of various applications. However, the most of existing studies through the three steps process (log collection, context inference, and service provision) did not consider smartphone applications (Apps) as the target service. Smartphone users still have to use Apps with manual controls by own decision. Therefore, in this paper, we propose a system to predict smartphone applications based on inferring user context. We define a mobile context model with a new level of context (Situation) and its inference method to perceive a user’s intention or purpose related to the App usage. Based on the Situation context, the system predicts Apps which can be useful and helpful for a user and automatically executes it on his/her smartphone. With the proposed system, it will be possible to autonomously provide and manage smartphone application services without users’ perception or intervention.     

基于双向长短时记忆联结时序分类和加权有限状态转换器的端到端中文语音识别系统

针对隐马尔可夫模型（HMM）在语音识别中存在的不合理条件假设,进一步研究循环神经网络的序列建模能力,提出了基于双向长短时记忆神经网络的声学模型构建方法,并将联结时序分类（CTC）训练准则成功地应用于该声学模型训练中,搭建出不依赖于隐马尔可夫模型的端到端中文语音识别系统;同时设计了基于加权有限状态转换器（WFST）的语音解码方法,有效解决了发音词典和语言模型难以融入解码过程的问题。与传统GMM-HMM系统和混合DNN-HMM系统对比,实验结果显示该端到端系统不仅明显降低了识别错误率,而且大幅提高了语音解码速度,表明了该声学模型可以有效地增强模型区分度和优化系统结构。     

A robust reliability prediction method using Weighted Least Square Support Vector Machine equipped with Chaos Modified Particle Swarm Optimization and Online Correcting Strategy

Accurate reliability prediction in engineering systems has drawn more and more attention over the past decades due to its important role in accessing the security condition of the system and providing safety operation basis, however, which still remains challenging caused by the mismatch of reliability prediction model especially for dynamic uncertainty of future sampled data. Therefore, a novel approach for reliability prediction with an optimal Online Correcting Strategy (OCS) combined with Weighted Least Square Support Vector Machine (WLSSVM) and Chaos Modified Particle Swarm Optimization (CMPSO) algorithm, named OCS–CMPSO–WLSSVM, are proposed, where WLSSVM models the functional relationship between input and output of the system, CMPSO optimizes the parameters of WLSSVM, and OCS modifies the model to reduce its mismatch as the system runs, respectively. The performance of the proposed model is demonstrated with five classic practical engineering examples and compared with the existing methods reported in literature in detail. The experimental results show that the proposed method not only has higher reliability prediction accuracy and robustness, but also has its superiority and applicability in other fields including time-ordered, feature-based regression problem and classification problem.     

A new platform for the prediction of field-dependent yield stress and plastic viscosity of magnetorheological fluids using particle swarm optimization

The yield stress and plastic viscosity of magnetorheological (MR) fluids are identified by fitting rheological models based on a selected dataset on a certain range of shear rates. However, the datasets are often arbitrarily determined as there is no standardized procedure available. To overcome this problem, a platform that capable to minimize the fitting error while considering the classification of the shear rate regions is needed. Therefore, this work proposed a new platform for the systematic prediction of field-dependent rheological characteristics using particle swarm optimization (PSO). PSO is a meta-heuristic algorithm for solving optimization problems based on a guided search of the defined problem space, which is governed by the objective function. An intersection point of low and high shear rate regions critical shear rate is formulated as part of the objective function to standardize the characterization within the defined regions. The objective function is inspired by the modified Bingham biplastic and Papanastasiou models to predict five magnetic field dependent-rheological parameters. In the development stage, the shear stress model was first established using a previously developed extreme learning machine method. Then, the codes of the PSO, objective functions and search space identification were developed and implemented. To validate the effectiveness of the proposed procedure, the platform performance was analysed at different algorithmic parameters and compared with the existing optimization methods. The simulation results indicated that the proposed platform performed better than the existing ones with R² of 0.943 and was able to systematically and accurately predict the rheological parameters.     

基于线性规划和遗传–粒子群算法的烧结配料多目标综合优化方法

针对钢铁企业二次配料工艺,本文采用将硫含量折算为可比成本,兼顾节能减排目标和配料成本,建立了二次配料多目标优化模型;提出了一种基于线性规划和遗传–粒子群算法（GA–PSO）的钢铁烧结配料优化方法.首先采用线性规划算法进行求解,若线性规划方法无法求得最优解,则采用GA–PSO算法进行搜索.该方法应用于某钢铁企业360m2生产线的＂配料优化与决策支持系统＂中,实际运行结果表明,该算法在保证烧结矿质量的前提下,能够有效地减少二氧化硫排放,降低配料成本.     

网络拓扑自动发现系统的设计与实现

在对目前的网络拓扑自动发现技术深入分析研究的基础上,设计并实现了一个网络拓扑发现系统。该系统能够从多个数据源中获取网络拓扑信息,具有网络拓扑自动分级发现和网络拓扑自动分层表示的功能。较传统的网络拓扑发现系统而言,该系统具有设计复杂度低、发现的网络拓扑完整和直观的优点。     

HEVC帧内预测Planar和DC模式的VLSI架构设计与实现

在研究新一代高性能视频编码标准（HEVC）帧内预测中planar和DC模式预测算法的基础上,分别设计了高效VLSI架构,通过状态机的自适应控制和模块的复用来实现速度的提高和面积的减少。针对planar模式,设计了一种基于状态机自适应控制的寄存器累加架构;针对DC模式,设计了一种基于算法的分割处理架构。实验结果表明,所设计的架构在TSMC180 nm的工艺下最高频率为350 MHz,面积合计为68.1 kgate,能够实现对4∶2∶0格式7 680×4 320@30 f/s视频序列的实时编码,最高工作频率可以达到23.4 MHz。     

基于时空特征向量的长短期记忆人工神经网络的城市公交旅行时间预测

针对"随着预测距离的增加,旅行时间预测的难度加大"的问题,提出了一种基于时空特征向量的长短期记忆(LSTM)和人工神经网络(ANN)的综合预测模型.首先,将24 h切分为288个时间切片,以生成时间特征向量;然后,基于时间切片建立LSTM时间窗口模型,该模型可解决长期预测的窗口移动问题;其次,将公交线路切分为多个空间切...     

联合MOD11A1和地面气象站点数据的多站点温度预测深度学习模型

针对地面气象站点分布稀疏影响站点间关系以及站点间的关系强度推理难的问题,提出一种基于联合MOD11A1和地面气象站点数据的多站点温度预测深度学习模型（GDM）。GDM包括时空注意力（TSA）、双向图神经长短期记忆（DG-LSTM）网络编码和边-点转换双向门控循环网络解码（EN-GRU）模块。首先使用TSA模块提取MOD11A1图像特征并形成多个虚拟气象站点的温度时间序列,缓解地面气象站点分布稀疏对站点间关系的影响;然后用DG-LSTM编码器通过融合两组温度时间序列来计算地面气象站点间和虚拟气象站点间的关系强度;最后用ENGRU解码器通过结合站点间的关系强度对地面气象站点的温度时间序列关系进行建模。实验结果表明,相较于二维卷积神经网络（2D-CNN）、长短期记忆全连接网络（LSTM-FC）、长短期记忆神经网络扩展网络（LSTME）和长短记忆与自适应提升集成网络（LSTM-AdaBoost）,GDM在10个地面气象站点24 h内温度预测的平均绝对误差（MAE）分别减小0.383℃、0.184℃、0.178℃和0.164℃,能提高未来24 h多个气象站点温度的预测精度。     

A parametric-based performance evaluation and design trade-offs for interconnect architectures using FPGAs for networks-on-chip

Network-on-Chip (NoC) interconnect fabrics are categorized according to trade-offs among latency, throughput, speed, and silicon area, and the correctness and performance of these fabrics in Field-Programmable Gate Array (FPGA) applications are assessed through experimentation and simulation. In this paper, we propose a consistent parametric method for evaluating the FPGA performance of three common on-chip interconnect architectures namely, the Mesh, Torus and Fat-tree architectures. We also investigate how NoC architectures are affected by interconnect and routing parameters, and demonstrate their flexibility and performance through FPGA synthesis and testing of 392 different NoC configurations. In this process, we found that the Flit Data Width (FDW) and Flit Buffer Depth (FBD) parameters have the heaviest impact on FPGA resources, and that these parameters, along with the number of Virtual Channels (VCs), significantly affect reassembly buffering and routing and logic requirements at NoC endpoints. Applying our evaluation technique to a detailed and flexible cycle accurate simulation, we drive the three NoC architectures using benign (Nearest Neighbor and Uniform) and adversarial (Tornado and Random Permutation) traffic patterns with different numbers of VCs, producing a set of load–delay curves. The results show that by strategically tuning the router and interconnect parameters, the Fat-tree network produces the best utilization of FPGA resources in terms of silicon area, clock frequency, critical path delays, network cost, saturation throughput, and latency, whereas the Mesh and Torus networks showed comparatively high resource costs and poor performance under adversarial traffic patterns. From our findings it is clear that the Fat-tree network proved to be more efficient in terms of FPGA resource utilization and is compliant with the current Xilinx FPGA devices. This approach will assist engineers and architects in establishing an early decision in the choice of right interconnects and router parameters for large and complex NoCs. We demonstrate that our approach substantially improves performance under a large variety of experimentation and simulation which confirm its suitability for real systems.     

Parallel protein secondary structure prediction schemes using Pthread and OpenMP over hyper-threading technology

Protein secondary structure prediction has a fundamental influence on today’s bioinformatics research. In this work, tertiary classifiers for the protein secondary structure prediction are implemented on Denoeux Belief Neural Network (DBNN) architecture. Hydrophobicity matrix, orthogonal matrix, BLOSUM62 matrix and PSSM matrix are experimented separately as the encoding schemes for DBNN. Hydrophobicity matrix, BLOSUM62 matrix and PSSM matrix are applied to DBNN architecture for the first time. The experimental results contribute to the design of new encoding schemes. Our accuracy of the tertiary classifier with PSSM encoding scheme reaches 72.01%, which is almost 10% better than the previous results obtained in 2003. Due to the time consuming task of training the neural networks, Pthread and OpenMP are employed to parallelize DBNN in the Hyper-Threading enabled Intel architecture. Speedup for 16 Pthreads is 4.9 and speedup for 16 OpenMP threads is 4 in the 4 processors shared memory architecture. Both speedup performance of OpenMP and Pthread is superior to that of other research. With the new parallel training algorithm, thousands of amino acids can be processed in reasonable amount of time. Our research also shows that Hyper-Threading technology for Intel architecture is efficient for parallel biological algorithms.

A context-augmented deep learning approach for worker trajectory prediction on unstructured and dynamic construction sites

Predicting workers’ trajectories on unstructured and dynamic construction sites is critical to workplace safety yet remains challenging. Existing prediction methods mainly rely on entity movement information but have not fully exploited the contextual information. This study proposes a context-augmented Long Short-Term Memory (LSTM) method, which integrates both individual movement and workplace contextual information (i.e., movements of neighboring entities, working group information, and potential destination information) into an LSTM network with an encoder-decoder architecture, to predict a sequence of target positions from a sequence of observations. The proposed context-augmented method is validated using construction videos and the prediction accuracy achieved is 8.51 pixels in terms of final displacement error (FDE), with an observation time of 3 s and prediction time of 5 s—5.4% smaller than using the position-based method. Compared to conventional one-step-ahead predictions, the proposed sequence-to-sequence method predicts trajectories over multiple steps to avoid error accumulation and effectively reduces the FDE by 70%. In addition, qualitative analysis is conducted to provide insights to select appropriate prediction methods given different construction scenarios. It was found that the context-aware model leads to better performance comparing to the position-based method when workers are conducting collaborative activities.     

基于遗传算法选优的集成手段与时序卷积网络的涡扇发动机剩余寿命预测

涡扇发动机作为航空航天领域的核心设备之一,其健康状况决定了航空器能否稳定可靠地运行。而对涡扇发动机的剩余寿命（RUL）进行判断,是设备监测与维护的重要一环。针对涡扇发动机监测过程中存在的工况复杂、监测数据多样、时间跨度长等特点,提出了一种遗传算法优选时序卷积网络（TCN）基模型的集成方法（GASEN-TCN）的涡扇发动机剩余寿命预测模型。首先,利用TCN捕获长跨度下的数据内在关系,从而对RUL作出预测;然后,应用GASEN集成多个独立的TCN,以增强模型的泛化性能;最后,在通用的商用模块化航空推进系统模拟模型（C-MAPSS）数据集上,对所提模型与当下流行的机器学习方法和其他的深度神经网络进行了比较。实验结果表明,在多种不同的运行模式和故障条件下,与流行的双向长短期记忆（Bi-LSTM）网络相比,所提模型都有着更高的预测准确率与更低的预测误差。以FD001数据集为例,在该数据集上所提模型的均方根误差（RMSE）相较Bi-LSTM低17.08%,相对准确率（Accuracy）相较Bi-LSTM高12.16%。所提模型在设备的智能检修与维护方面有着较好的应用前景。     

A study on the market instability index and risk warning levels in early warning system for economic crisis

Early warning system (EWS) can be treated as a pattern recognition problem since the distinctive feature of economic crisis makes it possible to distinguish critical and normal economic situations using a pattern classifier. Although the most works in EWS are mainly focused on training and pattern classifier, little attention has been paid to the effective indices or feature variables that allow closer look and analysis about the current instability nature of the economic crisis. This paper proposes to utilize market instability index (MII) and stepwise risk warning levels that can diagnose the current instability of the stock market to foretell how the current stock market will proceed in advance. This approach allows the proper policy actions to be taken for the possible financial crisis according to different risk warning levels of instability. Through empirical examples with Korean stock market and Greece stock market, the proposed method demonstrates its potential usefulness in an early warning system.     

湖库水质监测与水华预警信息系统

针对当前湖库水质监测及水华预测预警信息化发展相对落后的现状,开发一套集水质监测、水华预测预警功能于一体的智能化信息系统。采用Visual Studio 2010中的C++语言进行系统平台搭建,将网络通信、地理信息系统、SQL2005数据库等技术相结合,对湖库水质信息进行实时监测,并通过灰色-BP神经网络模型实现对湖库藻类水华较高精度的中长期预测预警的功能,为环保部门进行湖库水华防治提供有效的信息化决策平台。