Load-Balance Mechanism for WSNs Based on Minority Game with Dormancy Strategy

Load-balance is an important issue in wireless sensor networks (WSNs), especially in WSNs with hierarchical structure. Energy consumed unevenly will bring the production of hot spots. Hot spot will cause WSNs to divide some unconnected sub-networks and shorten the lifetime of WSNs. To tackle this problem, a load-balance mechanism is proposed based on minority game (MG) with dormancy strategy. This mechanism can cause the rich behaviors of cooperation , prolong lifetime of WSNs, and keep energy consumed evenly. This dormancy mechanism can save energy of nodes by keeping in sleep temperately . Simulation results show that the proposed strategy can efficiently enhance the lifetime of cluster and the lifetime of whole WSNs.

Mechanisms for the hiatus in global warming

The observed global mean temperature is the highest on record for the past decade but has plateaued to form an apparent"hiatus"in global temperature rise,with an almost zero short-term trend. Several speakers presented results on the hiatus and suggested possible mechanisms.

Pressure Angle in Parallel Mechanisms: From Planar to Spatial

This paper presents the evolution process of pressure angles from planar parallel mechanisms to spatial parallel mechanisms. Manipulability and condition number, which are frequently used in the optimum design of parallel mechanisms, are introduced from serial robots at first. Then, both theoretical analysis and practical experiences demonstrate that these concepts seem imperfect when they are used in parallel mechanisms. For this reason, this paper introduces the pressure angles in planar 4-bar mechanisms to spatial parallel mechanisms, which include redundant parallel mechanisms. Two kinds of pressure angles extracted from the determinant of direct and indirect Jacobian matrices are investigated. Moreover, two comprehensive and visible global performance indices are defined, showing the advantages in evaluating the workspace, singularity and motion/force transmission capabilities. With a 2-DOF planar and a 3-DOF spatial parallel mechanism as examples, the application of the performance indices is investigated and compared with the condition number at last. The proposed concept can be extended to other spatial parallel mechanisms.

New Institutional Mechanism in China Facilitating the Global Sustainability——Environment to Be Counted in Officials' Performance Rating

Having argued the importance of China's sustainable development in global sustainability, the authors review the achievements of China in sustainable development, especially its institutional construction. Environment to be counted in official's political performance rating system is thought of as a new institutional mechanism in China facilitating its sustainable development and then global sustainability. Then its significance is narrated and visions in future are envisioned. In the end, certain concrete suggestions for the rating system are given in a practical way.

Precipitation Distribution of the Extended Global Spring-Autumn Monsoon and Its Possible Formation Mechanism

The global monsoon(GM)comprises two major modes,namely,the solstitial mode and equinoctial asymmetric mode.In this paper,we extend the GM domain from the tropics to the global region and name it the global spring-autumn monsoon(GSAM),which mainly indicates a spring-autumn asymmetrical precipitation pattern exhibiting annual variation.Its distribution and possible formation mechanisms are also analyzed.The GSAM domain is mainly distributed over oceans,located both in the midlatitude and tropical regions of the Pacific and Atlantic.In the GSAM domains of both the Northern and Southern Hemispheres,more precipitation occurs in local autumn than in local spring.The formation mechanisms of GSAM precipitation vary according to the different domains.GSAM precipitation in the tropical domain of the Eastern Hemisphere is influenced by the circulation differences between the onset and retreat periods of the Asian summer monsoon,while tropical cyclone activities cause precipitation over the South China Sea(SCS)and western North Pacific(WNP).GSAM precipitation in the tropical domain of the Western Hemisphere is influenced by the tropical asymmetrical circulation between the Northern and Southern Hemispheres and the variation in the intertropical convergence zone(ITCZ)driven by the intensity of the sea surface temperature cold tongues over the equatorial eastern Pacific and eastern Atlantic.GSAM precipitation in the midlatitude domain is influenced by the differences in water vapor transportation and convergence between spring and autumn.In addition,GSAM precipitation is also affected by extratropical cyclone activities.

Preyminary Investigations on the Global Summer Monsoonal Drought Mechanics and the Latitudinal Teleconnections

The question of possible teleconnections between the middle latitude general circulation and the Indian south-west monsoon was investigated in this paper. Within the framework of a simple model it was shown that there can exist such an interaction via the ultra-long Rossby waves.

A Numerical Study of the Mechanism for the Effect of Northern Winter Arctic Ice Cover on the Global Short-Range Climate Evolution

By using a nine-layer global spectral model involving fuller parameterization of physical processes, with a rhomboidal truncation at wavenumber 15, experiments are performed in terms of two numerical schemes, one with long-term mean coverage of Arctic ice (Exp.1), the other without the ice (Exp.2). Results indicate that the Arctic region is a heat source in Exp.2 relative to the case in Exp.1. Under the influence of the polar heat source simulated, there still exist stationary wavetrains that produce WA-EUP and weak PNA patterns in Northern winter. That either the Arctic or the tropical heat source can cause identical climatic effects is due to the fact that the anomaly of the Arctic ice cover will directly induce a south-propagating wavetrain, and bring about the redistribution of the tropical heat source / sink. The redistribution is responsible for new wavetrains that will exert impact on the global climate. The simulation results bear out further that the polar region in Exp.2 as a heat source, can produce, by local forcing, a pair of positive and negative difference centers, which circle the Arctic moving eastwards. Observed in the Northern Hemisphere extratropics is a 40-50 day oscillation in relation to the moving pair, both having the same period.

Detailed Quantum Mechanical QSAR Analysis of Certain Aminopyrimidoisoquinolinequinones with Anticancer Activity

A detailed quantum mechanical analysis of electronic disposition of five aminopyrimidoisoquinolinequinones (APIQs) was performed after extraction of this subset of compounds from a larger data set of APIQs via a reported clustering methodology (Elfaki, et al. 2020). Both semi empirical PM3 method and DFT quantum mechanical methods were used to calculate global and local quantum mechanical descriptors (QMDs) to define the electronic environment of these molecules in attempt to rationalize their observed anti-cancer response variability. The biological response is the anticancer activity against human gastric adenocarcenoma (AGS) cell line. The correlation matrix between the calculated global electronic descriptors and biological activity demonstrated that the global dipole moment gives the highest correlation. The local electronic environment was analysed by The Mullikan charges (MC) and Fukui functions for N-5, C-6, C-8 in addition to the N atom of phenylamino side group at C-8. MCs furnished no useful information as each of these atoms had almost identical MC values for all the five compounds with exception of C-6 which gave varied values. Regressing MCs of C-6 against the response traces 60% of the latter variability. As C-6 is an extra annular methyl carbon adjacent to N-5 in isoquinoline residue of APIQ, we reasoned that the chemical reactivities of 4 out of the 5 APIQs might be due to a Chichibabin-type tautomerism implying a possible alkylation aspect in their mechanism of action. The corresponding Fukui functions (f-, f+ and f0) showed a considerable consistency with the patterns of chemical reactivity exhibited by this small set of APIQs.

一种可解释的云平台任务终止状态预测方法

基于特征选择和模型可解释方法构建可解释性强的云平台任务终止状态预测模型,该模型可视化任务/作业的静态和动态属性与终止状态之间的映射关系,进而找出负载特征与任务终止状态之间的映射机理.利用Google公开的工作负载监控日志,并加入云平台中任务的动态信息,采用沙普利加和解释(Shapley additive explain,SHAP)找出静态和动态属性对终止状态影响的重要性,利用变量重要性结合SHAP值和XGBoost模型,对任务终止状态预测模型建模后的结果进行解释,使用可视化技术呈现负载特征如何影响模型对不同任务终止状态的预测.用SHAP值绝对值的平均值衡量特征的重要性,实现任务不同终止状态特征重要性的全局可视化,根据结果筛选出对任务终止状态预测模型影响大的20个变量,作为特征筛选的依据;由可视化的结果可知,任务运行过程中,各特征的不同特征值对任务的终止状态有影响,不同特征值对终止状态的产生有不同的影响.特征选择结合模型可解释性方法运用于任务终止状态预测模型的构建流程中,可辅助构建高分类性能及易于理解的任务终止状态预测模型,通过对负载特征与任务终止状态之间映射机理的探索,可以优化云平台的调度机制.

天生全球化数字企业:概念内涵、理论基础与快速扩张机制

有别于传统天生全球化企业,天生全球化数字企业因独特的起源与技术能力,展现出差异化的国际化特征与快速扩张模式。对此,本文聚焦于数字时代的天生全球化现象,首先,对天生全球化数字企业的概念内涵进行界定,并与传统天生全球化企业进行比较研究,以深化对天生全球化新现象的理解。其次,基于内部化理论、过程理论、网络理论和LLL理论等国际商务理论,探讨数字时代天生全球化现象对已有理论应用的拓展。最后,以现有研究为基础,遵循“驱动因素—扩张模式—扩张结果”的逻辑,创新性地提炼出天生全球化数字企业的快速扩张机制,并聚焦探讨了基于用户规模和基于供给规模两种快速扩张模式。通过上述探索,期望能够为数字企业国际化现象提供一个更聚焦、更深入的阐释,并为后续研究提供相对明确且系统的研究指引。

关于设立面向全球科学研究基金的思考与建议

气候变化、健康危机和粮食安全等问题通常会超越国家间的地理边界,是典型的全球性挑战。随着不同挑战的日益加剧,设立全球科学研究基金已成为推动科技进步、探索创新解决方案和加强伙伴关系的重要手段。在此背景下,“研究设立面向全球的科学研究基金”是我国“十四五”时期(2021—2025年)完善科技创新体制的重要举措。为推进这一举措落实,本文在分析主要发达国家(美国、加拿大、英国、德国、日本、澳大利亚)和欧盟针对全球挑战的科研基金资助经验和广泛调研研讨的基础上,从人才项目、研究项目和大科学装置项目三个方面提出了我国未来设立面向全球的科学研究基金的建议,并探讨了相应的遴选、评估和管理机制,以期促进我国科研资助体系的创新发展,逐步提升我国在全球挑战研究中的国际影响力。

基于改进YOLOv5s的道路裂缝检测算法

为了解决道路巡检系统光学传感器采集的裂缝图像中颜色特征不明显且尺寸不规则造成检测精度不高、泛化能力不足的问题,提出改进YOLOv5s的裂缝检测算法。将结合深度可分离卷积(Depthwise Separable Convolution, DSC)的全局注意力(Global Attention Mechanism, GAM)引入主干特征提取网络,在降低注意力复杂度的同时获得丰富的跨维度特征,增强了裂缝的识别能力;采用空间金字塔软池化网络(Spatial Pyramid Softpool, SPSF),通过Softpool池化保留多维语义以减少信息弥散,提高了边界框回归的准确性;在颈部特征增强网络,运用空洞深度可分离卷积(Atrous DSC)进行下采样,通过扩大感受野加强深层和浅层信息的聚合能力,提高裂缝识别的泛化性。经过在自制道路裂缝数据集上的实验,相较于YOLOv5s,改进算法的mAP提高2.2%,有效提升了道路裂缝检测的准确性和对不同背景下裂缝识别的泛化能力。

基于双曲嵌入的露天矿区暗光环境下道路多目标检测模型

露天矿环境特殊,道路场景复杂多变,在光照不足时会导致矿区道路多目标识别不清、定位不准,进而影响检测效果,给矿区无人矿用卡车的安全行驶带来严重安全隐患。目前的道路障碍物检测模型不能有效解决矿区暗光环境对模型检测效果的影响,同时对矿区小目标障碍物的识别也有较大误差,不适用于矿区特殊环境下障碍物的检测与识别。针对上述问题,提出了一种基于双曲嵌入的露天矿区暗光环境下多目标检测模型。首先,在模型的图像预处理阶段引入卷积神经网路Retinex-Net对暗图像进行增强,提高图像清晰度;然后,针对数据集中特征过多而无重点偏好的问题,在加强特征提取部分添加全局注意力机制,聚集3个维度上更关键的特征信息;最后,在检测模型预测阶段引入双曲全连接层,以减少特征丢失,并防止过拟合现象。实验结果表明:(1)基于双曲嵌入的露天矿区暗光环境下道路多目标检测模型不仅对露天矿区暗光环境下的大尺度目标具有较高的分类与定位精度,对矿用卡车及较远距离的小尺度目标即行人也可准确检测与定位,能够满足无人矿用卡车在矿区特殊环境下驾驶的安全需求。(2)模型的检测准确率达98.6%,检测速度为51.52帧/s,较SSD、YOLOv4、YOLOv5、YOLOx、YOLOv7分别提高20.31%,18.51%,10.53%,8.39%,13.24%,对于矿区道路上的行人、矿用卡车及挖机的检测精度达97%以上。

改进YOLOv5s的蝗虫识别系统

为准确有效识别宁夏荒漠草原地区蝗虫种类,基于YOLOv5s网络模型提出一种复杂背景下蝗虫目标检测模型YOLOv5s-CG,在主干网络使用CoTNet保留复杂背景下的蝗虫特征信息,同时在颈部网络中融入GAM全局注意力机制提高检测模型的特征融合能力。结果表明,在对宁夏荒漠草原蝗虫进行识别时,模型YOLOv5s-CG精确率为92.5%,平均精度均值为93.2%,相比原始模型分别提高4.8个百分点和5.3个百分点,与Fast R-CNN、YOLOv3、YOLOv4、YOLOv5s、YOLOv6s、YOLOv7模型相比,YOLOv5s-CG网络模型对宁夏荒漠草原蝗虫具有更好的检测性能。基于YOLOv5s-CG,开发宁夏荒漠草原蝗虫识别APP系统,实现宁夏荒漠草原地区复杂背景下的蝗虫在线识别检测,为宁夏荒漠草原地区蝗虫的监管防控和综合治理提供数据支持。

局部注意力引导下的全局池化残差分类网络

大部分注意力机制虽然能增强图像特征,但没有考虑局部特征的关联性影响特征整体的问题。针对以上问题,本文提出局部注意力引导下的全局池化残差分类网络(MSLENet)。MSLENet的基线网络为ResNet34,首先改变首层结构,保留图像重要信息;其次提出多分割局部增强注意力机制(MSLE)模块,MSLE模块将图像整体分割成多个小图像,增强每个小图像的局部特征,通过特征组交互的方式将局部重要特征引导到全局特征中;最后提出池化残差(PR)模块来处理ResNet残差结构丢失信息的问题,提高各层之间的信息利用率。实验结果表明,MSLENet通过增强局部特征的关联性,在多个数据集上均有良好的效果,有效地提高了网络的表达能力。

浙江新材料科创高地支撑全球先进制造业基地建设的动力机制与实现路径

新材料是浙江省“315”科技创新体系三大科创高地之一,是支撑浙江建设全球先进制造业基地着重打造的4个世界级先进产业群之一。依据浙江省建设全球新材料先进制造业基地的目标与要求,构建由新材料科创高地为基本支撑,以“政、产、学、研、用”多元创新主体协同,以“制度、技术、人才、市场、资本”多要素协同为基本动力的新材料全球先进制造业基地建设动力机制模型,并据此提出加强新材料产业集群顶层设计、支持新材料产业研发创新、完善新材料产业集群价值网络、健全专精特新小巨人企业梯度培育机制的实现路径,为新形势下浙江新材料产业发展和新材料先进制造业基地建设提供借鉴。

基于包络学习和分级结构一致性机制的不平衡集成算法

集成方法是不平衡学习方法的重要分支,然而,现有不平衡集成方法均作用于原样本而没考虑样本的结构信息,因此其效能仍然有限.样本的结构信息包括局部和全局结构信息.为了解决上述问题,本文提出了一种基于深度样本包络网络(Deep Instance Envelope Network,DIEN)和分级结构一致性机制(Hierarchical Structure Consistency Mechanism,HSCM)的不平衡集成学习算法.该算法在考虑局部流形和全局结构信息的情况下,通过多层样本聚类,生成高质量的多层包络样本,从而实现类平衡化.首先,算法基于样本近邻拼接和模糊C均值聚类算法,设计DIEN来挖掘样本的结构信息,得到深度包络样本.然后,设计局部流形结构度量和全局结构分布度量来构建HSCM用于增强层间样本的分布一致性.接着,将DIEN和HSCM结合起来,构建出优化后的深度样本包络网络——DH(DIEN with HSCM).之后,将基分类器应用于包络样本.最后,设计bagging集成学习机制来融合基分类器的预测结果.文末组织了多组实验,采用了十多个公共数据集和有代表性的相关算法进行验证比较.实验结果表明,本文算法在AUC(Area Under Curve),F-measure等四个性能指标上显著最优.

融合边界注意力的特征挖掘息肉小目标网络

从结肠图像中分割息肉小目标病变区域对于预防结直肠癌至关重要,它可以为结直肠癌的诊断提供有价值的信息。然而目前现有的方法存在2个局限性:一是不能稳健捕获全局上下文信息,二是未能充分挖掘细粒度细节特征信息。因此,提出融合边界注意力的特征挖掘息肉小目标网络(transformer feature boundary network,TFB-Net)。该网络主要包括3个核心模块:首先,采用Transformer辅助编码器建立长程依赖关系,补充全局信息;其次,设计特征挖掘模块进一步细化特征,学习到更好的特征;最后,使用边界反转注意力模块加强对边界语义空间的关注,提高区域辨别能力。在5个息肉小目标数据集上进行广泛实验,实验结果表明TFBNet具有优越的分割性能。

多视野精细分析下的弱监督目标定位算法

针对多尺度目标定位精度较差,难以捕获完整目标边界的问题,设计一种多视野精细分析模块并融入通道与空间注意力机制抑制背景噪声的干扰,获取多尺度目标的高分辨率特征。利用随机特征选取模块获取特征图随机位置的组合,聚合多个位置图获取最具辨别性的位置及其它位置的信息,融合浅层生成的类激活图与聚合类激活图获取细粒度位置信息,捕获完整的目标边界。与现有的弱监督定位方法相比,在解决多尺度目标定位效果差和局部最优问题上具有一定的优势。

GFENet:基于Transformer的高效医学图像分割网络

医学图像分割技术在疾病诊断中发挥着重要作用,针对传统网络分割模型中存在参数量大、网络计算效率低等问题,本文提出名为GFENet的高性能低复杂度分割网络,GFENet以金字塔结构的视觉Transformer作为网络主干提取图像特征,分别采用线性预测模块和渐进式局域解码器对提取出来的低级特征和高级特征进行局部增强处理,最后采用GFE模块中的全局注意力机制将位置信息嵌入高级特征.在Kvasir和CVC-ClinicDB数据集上进行实验验证,GFENet的mDice分别为94.1%和94.6%;其在CVC-ColonDB和ETIS数据集上的泛化性能优于其他对比模型.为验证模型低复杂度的同时保持高分割准确率,本文将GFENet与现有的高性能模型和轻量化模型进行对比,GFENet在CVC-ColonDB数据集上以81.5%的mDice高于其它对比模型,以23.1M的参数量优于其他高性能分割模型.