基于邻居决策的协同过滤推荐算法

协同过滤技术应用于个性化推荐系统中,稀疏性问题和可扩展性问题成为亟需解决的问题。针对传统方法的不足,提出一种凭借邻居数做决策的方法,比较各个待测位置的用户邻居数和项目邻居数,由数量多的一方作预测,同时对预测值判定给出一种合理而有效的度量方法。实验结果表明,该方法能够提高推荐质量。     

Two-Parameter Thermal Lattice BGK Model with a Controllable Prandtl Number

In this paper, two time relaxation parameters are introduce to a thermal lattice BGK, model to make its Prandtl number controllable. The dependency of the Prandtl number on the two parameters is derived. Numerical measurement of the transport coefficients is used to demonstrate the validity of the method. Furthermore, two examples of convective heat transfer are calculated, with one to show the effectiveness, and the other to show the breakdown of the two-parameter formulation under different conditions.     

数码相机面壳注射成型的浇口优化设计

通过运用Moldnow软件对数码相机面壳注射成型进行了流动模拟分析,预测在不同浇口位置和数目的情况下可能存在的气泡和熔接痕位置,确定了最佳的浇口数目和位置.     

一种改进的ALOHA防碰撞算法

针对现有动态帧时隙ALOHA防碰撞算法在射频识别系统中的标签识别效率最高只能达36.8%的问题,利用标签码元序列的唯一性,改进经典的动态帧时隙ALOHA防碰撞算法中随机选择时隙的方式,提出一种基于标签码元的碰撞序列进行时隙选择的方法,有效降低了标签碰撞的概率,从而提高系统识别效率。仿真结果表明改进的ALOHA标签防碰撞算法识别效率最低为37.5%,随着实际标签数目与碰撞位序列所能确定的标签数目越接近,识别效率越高,最高能达到100%,明显优于现有的动态帧时隙ALOHA算法。     

交换机系统割接方案

在全国电力系统的语音调度专网通信系统中,由于系统发展的需要,经常涉及到电话交换机的移机及割接。由于调度通信系统的重要性及特殊性,在交换机移机及割接过程中要保持通信的畅通,不能中断。在一些核心的大局的割接过程中,在业务复杂繁多的情况下,如何保证通信不中断,做到平滑、无缝的割接,是许多大局都面临的问题。详细介绍一种调度交换机割接并移机的方案,可以提供一些这方面的参考。     

单、双层不等匝正弦绕组线模的尺寸计算

为了减少谐波磁势,削弱同步附加转矩,提高电机的效率及改善电机性能,三相异步电机的定子线圈广泛采用单、双层不等匝正弦绕组(以下简称正弦绕组)。由于正弦绕组每极每相下各线圈的跨距和导体数均不相等,其绕线模尺寸计算复杂,文章介绍了它的计算方法。     

Balancing CapEx reduction and network stability with stable routing–virtual topology capacity adjustment (SR–VTCA)

This paper investigates the merits of the SR–VTCA (stable routing–virtual topology capacity adjustment) approach as a mechanism to find a beneficial trade-off between network stability and reduction in capital expenditures (CapEx). These are two main objectives for the entities that own the optical infrastructure, such as network operators (NOs), and those also acting as Internet service providers (ISPs). The SR–VTCA scheme is a novel approach to adapt transparent optical networks to time-varying traffic by adjusting the number of lightpaths between node pairs, while keeping the IP routing unchanged. Lightpath bundling (LB) and anycast (AS) switching are combined in SR–VTCA operation to advertise lightpath additions/removals to the IP layer as mere adjustments (increments or decrements) in the capacity, allowing to keep the IP routing stable, and thus, simplifying control plane operations. On the contrary, a fully-reconfigurable (FR) network design, where IP routing can be also modified, would increase the burden in the control plane, but at a higher CapEx reduction, since the optical infrastructure is used more efficiently. In this work, we investigate the CapEx overprovision introduced by SR–VTCA with respect to a FR scheme. In order to do this, SR–VTCA planning problem is first modeled as a MILP formulation. A heuristic procedure based on traffic domination is then proposed to solve large instances of the problem. Exhaustive experiments are conducted comparing the SR–VTCA solutions obtained by the aforementioned MILP and heuristic proposal with solutions found by other optimization methods presented in the literature to solve the FR planning problem. Finally, the results show that SR–VTCA can achieve similar results to the FR case in terms of CapEx reduction, while a huge number of IP reroutings are saved by maintaining IP stability. Thus, SR–VTCA provides an advantageous balance between CapEx overprovisioning and the control plane overhead associated with IP rerouting.     

T-NTRU物联网动态接入认证技术

物联网终端安全接入认证是保证电力物联网大规模建设的关键性技术.传统认证方案通常采用的椭圆曲线密码算法,计算量大,不能抵抗量子攻击.数论研究单元算法可以抵抗量子攻击,与椭圆曲线密码采用点乘算法相比,计算速度快.因此提出了一种时间变换数论研究单元算法物联网动态安全接入认证算法,使用动态变化时间序列通过哈希函数产生动态密钥,...     

Generation and Identification of the Number of Copies of Exogenous Genes and the T-DNA Insertion Site in SCN-Resistance Transformation Event ZHs1-2

Soybean cyst nematode (SCN, Heterodera glycines Ichinohe) causes an estimated economic loss of about USD 3 billion each year in soybean (Glycine max L.) production worldwide. Overexpression of resistance genes against SCN provides a powerful approach to develop SCN resistance cultivars in soybean. The clarification of molecular characterization in transformation events is a prerequisite for ecological risk assessment, food safety, and commercial release of genetically modified crops. Here, we generated transgenic events harboring the BCN (beet cyst nematode) resistance Hs1^pro−1 gene using the Agrobacterium-mediated method in soybean, evaluated their resistance to SCN infection, and clarified the molecular characterization of one of the transformation events. Five independent and stable inheritable transformation events were generated by an Agrobacterium-mediated transformation method. SCN resistance tests showed the average number of developed females per plant and female index (FI) in T4 ZHs1-1, ZHs1-2, ZHs1-3, ZHs1-4, and ZHs1-5 transformation events were significantly lower than that in the nontransgenic control. Among these, the ZHs1-2 transformation event had the lowest number of developed females per plant and FI. Southern hybridization showed the exogenous target Hs1^pro−1 gene was inserted in one copy and the Bar gene was inserted two copies in the ZHs1-2 transformation event. The exogenous T-DNA fragment was integrated in the reverse position of Chr02: 5351566–5231578 (mainly the Bar gene expression cassette) and in the forward position of Chr03: 17083358–17083400 (intact T-DNA, including Hs1^pro−1 and Bar gene expression cassette) using a whole genome sequencing method (WGS). The results of WGS method and Southern hybridization were consistent. All the functional elements of exogenous T-DNA fragments were verified by PCR using specific primer pairs in the T5 and T6 ZHs1-2 transformation events. These results demonstrated that the overexpression of Hs1^pro−1 gene enhanced SCN resistance, and provide an important reference for the biosafety assessment and the labeling detection in transformation event ZHs1-2.     

基于深度信任网络模型的乌东选煤厂铁路车号图像识别方法

精准识别铁路车号可以为煤厂装车提供依据,从而保证装车环节高效顺利地完成。为此,提出了基于深度信任网络模型的乌东选煤厂铁路车号图像识别方法。首先,利用高速摄像机设备采集原始的车号图像,并利用索贝尔算子检测图像边界;然后,根据列车车号的字体笔画宽度特点,采笔画宽度变换算法定位确定图像中的车号区域,并利用LBP算法提取车号区域内的特征;最后,将提取的特征输入到深度信任网络模型中,在训练网络模型并不断更新参数后,准确识别车号图像。实验表明：该方法能够精准识别乌东选煤厂铁路列车车号图像。在深度信任网络模型中,当受限玻尔兹曼机网络为4层、隐含层节点个数为128个时,该模型的分类识别能力最强,训练损失最小,性能最佳。