File and workload allocation on a local multi-access computer network: incorporating local processing and communication overhead

The paper presents an analytical model to allocate files and workload simultaneously among heterogeneous servers for a distributed database where the servers are connected by a local multi-access computer network. The allocation problem is presented in the form of a nonlinear integer programming model. The model incorporates both local processing and communication overhead. The problem is proved to be NP-complete and thus a heuristic is developed. To illustrate its suitability, the heuristic is compared with an exhaustive enumeration method for small problems and with a genetic algorithm for large problems. Computational results demonstrate that the proposed model and heuristic provide effective database designs     

Performance improvement using time delays in multivariable controller design

A new controller for linear multivariable ordinary systems is suggested in which distributed delays are included in the feedback loop. A general design approach to obtain this type of controller is suggested. It is shown that the resulting closed-loop system is asymptotically stable under some sufficient conditions. Sufficient conditions are derived under which the disturbance attenuation, the robustness against parameter variations, and the time-delay stability margins are improved by the proposed controller. The state feedback tracking controller and the dual-state observer are obtained by modification of the proposed controller.     

基于自动编码器的深度伪造图像检测方法

基于深度学习的图像伪造方法生成的图像肉眼难辨,一旦该技术被滥用于制作虚假图像和视频,可能会对国家政治、经济、文化造成严重的负面影响,也可能会对社会生活和个人隐私构成威胁。针对上述问题,提出了一种基于自动编码器的深度伪造Deepfake图像检测方法。首先,借助高斯滤波对图像进行预处理,提取高频信息作为模型输入;然后,利用自动编码器对图像进行特征提取,并在编码器中添加注意力机制模块以获取更好的分类效果;最后,通过消融实验证明,采用所提的预处理方法和添加注意力机制模块有助于伪造图像检测。实验结果表明,与ResNet50、Xception以及InceptionV3相比,所提方法在数据集样本量较小且包含的场景丰富时,可以有效检测多种生成方法所伪造的图像,其平均准确率可达97.10%,明显优于对比方法,且其泛化性能也明显优于对比方法。     

基于朴素贝叶斯算法的中文评论分类

针对中文评论分类问题,采用朴素贝叶斯算法进行深入研究.首先,根据中文评论分类的需求设计了朴素贝叶斯分类器;然后,使用WEKA以不同特征提取方式对其功能性进行了对比分析.通过一系列的实验数据的横向对比表明,在朴素贝叶斯分类器下采用集成特征选取时文本分类的准确率最佳,准确率达97.65％,验证了朴素贝叶斯分类器在处理中文评...     

基于LASSO回归模型的网贷借款成功影响因素挖掘

随着P2P网络借贷交易量的增大,对P2P交易数据的挖掘和分析备受关注,其中一项重要的研究课题是网络借款成功率的影响因素分析.现有的文献多采用线性回归方法对该课题进行研究,但未考虑变量之间的多重共线性和采用最优变量子集建立回归模型的问题.本文采用Lasso回归方法,建立最优变量子集的回归模型对影响网络借款成功率的因素进行分析,避免了多重共线性问题对模型的干扰,同时提高了模型对数据的拟合精度.对Lending Club平台的借贷数据的实证分析结果显示,本文方法在模型的拟合精度和避免共线性方面优于对比方法.     

Development of a simple model for batch and boundary information updation for a similar ship’s block model

In early 2000,large domestic shipyards introduced shipbuilding 3D computer-aided design (CAD) to the hull production design process to define manufacturing and assembly information.The production design process accounts for most of the man-hours (M/H) of the entire design process and is closely connected to yard production because designs must take into account the production schedule of the shipyard,the current state of the dock needed to mount the ship’s block,and supply information.Therefore,many shipyards are investigating the complete automation of the production design process to reduce the M/H for designers.However,these problems are still currently unresolved,and a clear direction is needed for research on the automatic design base of manufacturing rules,batches reflecting changed building specifications,batch updates of boundary information for hull members,and management of the hull model change history to automate the production design process.In this study,a process was developed to aid production design engineers in designing a new ship’s hull block model from that of a similar ship previously built,based on AVEVA Marine.An automation system that uses the similar ship’s hull block model is proposed to reduce M/H and human errors by the production design engineer.First,scheme files holding important information were constructed in a database to automatically update hull block model modifications.Second,for batch updates,the database’s table,including building specifications and the referential integrity of a relational database were compared.In particular,this study focused on reflecting the frequent modification of building specifications and regeneration of boundary information of the adjacent panel due to changes in a specific panel.Third,the rollback function is proposed in which the database (DB) is used to return to the previously designed panels.     

A knowledge-based approach to safety evaluation for plant start-up

To improve the safety of plant start-up operation, a safety evaluation system has been developed. As a key component in an operational design support system, the evaluation system examines any potential hazards during start-up operation simulation. The evaluation system is integrated into an operational design methodology which designs operable processes by proposing alternatives, examining process safety and operability, and modifying operating procedures or plant structures. Issues for both methodology and implementation of a G2-based expert system are discussed. Finally, the system is applied to an industrial hydrodesulphurization process.     

固体氧化物燃料电池离子- 电子混合传导性Ni/ScSZ 燃料极的制备以及性能分析

在薄膜固体氧化物燃料电池中,为增加发生电化学反应的三相界面面积,燃料极和空气极通常制备成离子-电子混合传导性质的多孔形态。该研究将燃料极电化学反应(氢氧化反应)的催化剂——镍(Ni)和氧离子传导性好的氧化钪稳定的氧化锆(ScSZ)混合制备成离子-电子混合传导燃料极。实验结果显示,与纯 Ni 燃料极相比,所制备的 Ni/ScSZ 燃料极因其三相界面面积的增加,可显著降低极化电阻,但同时欧姆电阻也显著增加,从而最终导致燃料电池单池性能总体下降。