高pH条件纳米Fe_3O_4催化H_2O_2分解去除水中4-氯酚

为了克服非均相fenton催化反应中溶液初始pH过低的缺点,开展了pH=5时Fe3O4/H2O2体系催化氧化去除水中4-氯酚实验。试验结果表明:制备的铁氧化物为纯Fe3O4,具有反尖晶石八面体结构。纳米Fe3O4能够高效催化H2O2分解氧化水中的4-氯酚,反应180 min后4-氯酚的去除率达到96.8%。当催化剂投量小于2.0 g/L,4-氯酚的去除率随着催化剂投量的增加而升高;当催化剂投量大于2.0 g/L,去除率会随着投量的增加而降低。随着H2O2投量的增加,4-氯酚的去除率先升高后降低。溶液的初始pH对4-氯酚的去除影响较大,pH越低,去除速率和效率越高。重复使用5次以后,纳米Fe3O4仍然保持较高的催化活性。     

基于差分进化算法的铁磁物体磁化率优化方法

针对工程中铁磁物体磁化率难以确定的问题,以单元表面积分铁磁物体感应磁场模型为基础,利用磁场测量值建立了多维磁化率数学优化模型,并运用具有全局多维寻优能力的进化差分算法对多维磁化率数学模型进行了求解,进而确定了铁磁物体的感应磁场分布。最后,设计了空心圆筒磁测实验,实验结果表明:该方法能有效获取铁磁物质多维磁化率,还能精确计算铁磁物体的感应磁场分布。     

The Production and Deposition of the Guisborough Helmet

In September 1878 the British Museum received a donation of a folded metal plate which had been embossed and gilded. Restoration of the object revealed that the folded plate was in fact the outer casing of a Roman copper alloy cavalry sports type helmet which dates to the late second or third century AD. The story of what became known as the Guisborough helmet began 14 years earlier in what was then the North Riding of Yorkshire when workers employed by the Cleveland Railway Company discovered a ‘very curious plate of metal’ during road construction. It appears that the helmet was deliberately prepared for deposition and buried at a depth of c30 cm close to a stream. Examination by the authors suggests that this was a high value item that had been repaired many times. Experimental evidence suggests that the decoration alone required approximately 16 h to fashion.     

机动目标IMM三维并行滤波跟踪算法

由于复杂的空中目标机动,其三维方向的机动强度是不一致的,传统IMM算法存在模型匹配不准确的问题,提出一种机动目标IMM三维并行滤波的跟踪算法。算法以CV和修正的CS模型为子集,在3个坐标轴上分别根据目标机动的分量实际更新其模型概率,并行IMM滤波方法,尽量确保模型的适配性,提高滤波精度。仿真结果表明,该算法比传统IMM方法跟踪精度更高,对空中机动目标跟踪适应性更强。     

COSMIC 掩星反演的水汽廓线质量分析

利用2006—2012年北半球冬季低纬度地区(30°S~30°N)无线电探空站数据及全球大气成分和气候监测再分析数据对1000 h Pa~200 h Pa高度层的气象、电离层与气候星座观测系统全球定位系统掩星反演的比湿廓线进行了精度和可靠性验证。结果表明,水汽对气象、电离层与气候星座观测系统掩星反演影响较大,尤其在中、低对流层及热带地区等水汽含量比较大的地区,且气象、电离层与气候星座观测系统掩星数据在850 hpa以下可能并不太适用于评估其他数据。     

Preemptive parallel‐machine scheduling with a common server to minimize makespan

We consider parallel‐machine scheduling with a common server and job preemption to minimize the makespan. While the non‐preemptive version of the problem is strongly NP‐hard, the complexity status of the preemptive version has remained open. We show that the preemptive version is NP‐hard even if there is a fixed number of machines. We give a pseudo‐polynomial time algorithm to solve the case with two machines. We show that the case with an arbitrary number of machines is unary NP‐hard, analyze the performance ratios of some natural heuristic algorithms, and present several solvable special cases. © 2017 Wiley Periodicals, Inc. Naval Research Logistics 64: 388–398, 2017     

From Nationalism to Factionalism: Faultlines in the Naga Insurgency

Do ceasefires or peace talks create fragmentation in the insurgent groups? Rather than proposing claims that can offer predictions about armed groups behaviour under ceasefires or peace processes, the analysts tend to focus largely on the dynamics between state and non-state actor. The experts pay little attention to overtime changes in social and local political context which might contribute to propelling a rebel group towards fragmentation and factionalism. The present study intends to fill this gap by exploring the shifting role of public opinion and ethnic support for the peace talks to ascertain whether it can increase the likelihood of factionalism in rebel groups or not. This article applies this approach to the case of Naga National Movement (1947–2015) in India, and finds that the proposed variable appears to have increased the frequency of factionalism in the movement.     

稀疏卷积计算高效数据加载与输出缓存策略

针对现有神经网络加速器在处理稀疏神经网络时存在的数据加载效率低、乘加资源利用率低、输出缓存寻址逻辑复杂等问题,提出了稀疏卷积计算高效数据加载与输出缓存策略。将属于同一输入通道的非零输入特征图像数据和非零权重进行全对全乘累加运算,降低了非零数据配对难度,提高了乘加资源利用率;通过采用输入驻留计算,以及密集型循环加载特征图像数据,大幅减少了数据片外调取次数;优化了输出缓存设计,解决了现有方案中存在的输出缓存地址访问争用、存储拥塞等问题。实验表明,与采用类似架构的细粒度脉动加速器相比,在处理单元面积上减少了21.45%;在数据加载速度方面平均提高了117.71%在平均乘法器利用率方面提高了11.25%,达到89%。     

多核贝叶斯优化的模型决策树算法

构造模型决策树时超参数较多,参数组合复杂,利用网格搜索等调参方法将会消耗大量的时间,影响模型性能的提升。提出了一种多核贝叶斯优化的模型决策树算法,该算法为应对不同分类数据特性,采用三种高斯过程建模寻优,利用贝叶斯优化技术,选出最优的参数组合。实验结果表明,所提算法在参数寻优上要优于传统的模型决策树寻优方法,并且能够在迭代次数不多的情况下找到全局最优参数值,在一定程度上提升了算法的分类性能,节省了大量的调参时间。     

融合空间偏好和语义的个体活动识别方法

个体活动识别对用户画像、个性化推荐、异常行为检测、群体行为分析和基于活动的资源配置优化具有重要价值。提出了一种基于稀疏的社交媒体签到数据的个体活动语义识别方法,从签到数据中提取活动行为的时间周期性和趋势性特征,并采用空间偏好量化算法,从个体与群体活动的空间关联中提取群体和个体的空间访问偏好,使用自然语言嵌入工具BERT模型提取访问兴趣点的语义。时间特征、空间偏好特征和访问兴趣点名称语义特征共同构成表征群体、个体偏好的时空联合特征,通过极限梯度提升分类器对其进行分类,得到活动语义识别结果。在Foursquare数据集上的对比实验和消融实验中验证了所提活动语义识别模型可以有效提升活动语义识别的准确性。