Temporal evolution of contacts and communities in networks of face-to-face human interactions

Temporal dynamics of social interaction networks as well as the analysis of communities are key aspects to gain a better understanding of the involved processes, important influence factors, their effects, and their structural implications. In this article, we analyze temporal dynamics of contacts and the evolution of communities in networks of face-to-face proximity. As our application context, we consider four scientific coil- ferences. On a structural level, we focus on static and dynamic properties of the contact graphs. Also, we analyze the resulting community structure using state-of-the-art automatic community detection algorithms. Specifically, we analyze tile evolution of contacts and communities over time to consider the stability of the respective conmmnities. Furthermore, we assess different factors which have an influence on the quality of com- munity prediction. Overall, we provide first important insights into the evolution of contacts and communities in face-to-face contact networks.     

融合多头自注意力的问答社区专家推荐算法

专家推荐是在线问答社区的研究热点之一,但现有的算法大多关注用户的静态兴趣和问题信息的匹配,忽视了对用户的动态兴趣表征信息的有效捕捉,从而导致推荐的准确度不足。针对上述问题,提出了融合多头自注意力的问答社区专家推荐算法。首先,构造由卷积神经网络和注意力机制组成的问题编码器,来处理目标问题和用户历史回答问题,提取对应的问题表征;其次,将用户历史回答问题序列当作时间序列,利用多头自注意力机制学习序列中所蕴涵的动态兴趣表征,结合用户的静态兴趣表征,获取用户的综合兴趣表征;最后,将目标问题表征和用户综合表征进行相似性计算产生推荐结果。利用来自知乎问答社区的真实数据进行了不同参数配置及不同算法的对比实验,实验结果表明该算法性能要明显优于目前较流行的深度学习专家推荐算法。     

Influence of Phenolic acids on microbial populations in the rhizosphere of cucumber

Experiments were conducted to determine whether changes in soil microbial populations that occur in response to additions of certain allelopathic phenolic acids to bulk soil also occur in the rhizosphere. Cucumber seedlings were transplanted into cups containing a nutrient-enriched mixture of Portsmouth B₁, soil and sand and were watered five times (once every 48 hr) with aqueous solutions of ferulic,p-coumaric, or vanillic acid (each at 0, 0.25, or 0.50ol/g soil material). Nutrient solution was applied on alternate days. Leaf growth was suppressed by up to 42% by phenolic acids, but changes in root growth varied with the compound and concentration in solution. Significant increases (over 600% relative to controls) in populations of fast-growing bacteria in the rhizosphere were detected after two but not after five treatments, and increases (400% relative to controls) in numbers of fungal propagules were detected after five treatments. Such increases suggested that chronic exposure to a phenolic acid might resuit in high populations of rhizosphere microorganisms that could metabolize the compounds and thus alter observable responses by the plant. To test this, plants were watered repeatedly with a low-concentration solution of ferulic acid (chronic treatments; 0.0 or 0.1mol/g soil material in one experiment, 0.000 or 0.025imol/g soil material in a second) and then once with a highconcentration solution (acute treatment; 0.0, 0.5, or 1.0mol/g soil material in the first experiment; 0.000, 0.125, or 0.250mol/g soil material in the second).Paper No. 12385 of the Journal Series of the North Carolina Agricultural Research Service, Raleigh, North Carolina. The use of trade names in this publication does not imply endorsement by the North Carolina Agricultural Research Service or the United States Department of Agriculture of the products named, nor criticism of similar ones not mentioned. Acute treatments and some chronic treatments suppressed leaf growth, but results were inconsistent for root growth. Acute treatments increased numbers of several types of bacteria in the rhizosphere but had inconsistent effects on fungi. Chronic treatments had no effect on numbers of bacteria or fungal propagules in the rhizosphere. Furthermore, chronic treatments did not alter responses of plants or microbial populations to the subsequent acute treatment. Results demonstrated that phenolic acids in soil, which must pass through the rhizosphere before interaction with plant roots can occur, alter the microbial ecology of the rhizosphere. However, microbially mediated acclimation of plants to relatively high concentrations of ferulic acid was not observed.     

基于梯度的重叠式层次社区检测

社区检测(community detection)任务一直是数据挖掘领域的一个研究热点,近年来,深度学习和图链接数据呈现出多样化和复杂化的发展趋势,层次(Hierarchical)社区检测逐渐成为研究的焦点.层次社区检测任务的目标是,在将同质图中相似的节点聚集到社区中的同时,学习社区之间的层次结构关系,以更好的理解图数据结构.社区间层次关系的引入给社区检测算法带来了更复杂的建模挑战.针对该任务,已经有一些有效的启发式的方法被提出,但是受限于社区分布形态的简单假设和离散的优化学习方式,它们无法描述更复杂的图链路数据,也无法和其它有效的连续优化算法组合获得更好的结果.为了解决这个问题,本文首次尝试建模复杂的重叠式(overlapping)层次社区结构,提出简洁的节点嵌入和社区检测双任务优化模型,通过梯度更新的方式来灵活地探索节点和重叠式层次社区的隶属关系.在学习过程中,我们可以分别获得节点和社区的嵌入表示,以应用于丰富的下游任务.     

基于迁移学习的卷积神经网络通道剪枝

卷积神经网络在计算机视觉等多个领域应用广泛,然而其模型参数量众多、计算开销庞大,导致许多边缘设备无法满足其存储与计算资源要求。针对其边缘部署困难,提出使用迁移学习策略改进基于BN层缩放因子通道剪枝方法的稀疏化过程。本文对比不同层级迁移方案对稀疏化效果与通道剪枝选取容限的影响;并基于网络结构搜索观点设计实验,探究其精度保持极限与迭代结构的收敛性。实验结果表明,对比原模型,采用迁移学习的通道剪枝算法,在精度损失不超过0.10的前提下,参数量减少89.1%,模型存储大小压缩89.3%;对比原剪枝方法,将剪枝阈值从0.85提升到0.97,进一步减少参数42.6%。实验证明,引入迁移策略更易实现充分的稀疏化,提高通道剪枝阈值选取容限,实现更高压缩率;并在迭代剪枝的网络结构搜索过程中,提供更高效的搜索起点,利于快速迭代趋近至搜索空间的一个网络结构局部最优解。     

长江大保护与流域生态学

长江经济带发展的基本点是"共抓大保护、不搞大开发"。在梳理了长江流域主要生态环境问题基础上,从长江大保护的核心是统筹山水林田湖草、山水林田湖草生命共同体即是流域生态系统、流域生态系统是典型的集合生态系统的角度,论述了长江大保护与流域生态学的关系,并从科学研究、规划管理及教育宣传等方面展开思考提出建议,为实现长江大保护的国家需求提供服务。     

Influence of a high‐head dam as a dispersal barrier to fish community structure of the Upper Mississippi River

In river systems, high‐head dams may increase the distance‐decay of fish community similarity by creating nearly impermeable dispersal barriers to certain species from upstream reaches. Substantial evidence suggests that migratory species are impacted by dams, and most previous studies in stream/river networks have focused on small streams and headwaters. Here, we assess whether a high‐head dam (Lock and Dam 19; LD 19) on a large river, the Upper Mississippi River (UMR), substantially alters fish community structure relative to variability expected to occur independent of the dam's effect as a fish dispersal barrier. Using fish catch per unit effort data, we modelled the distance‐decay function for the UMR fish community and then estimated the similarity that would be expected to occur across LD19 and compared it with measured similarity. Measured similarity in the fish community above and below LD19 was close to the expected value based on the distance‐decay function, suggesting LD19 does not create an abrupt transition in the fish community. Although some migratory fish species no longer occur above LD19 (e.g., skipjack herring, Alosa chrysochloris), these species do not occur in high abundance below the dam and so do not drive variation in fish community structure. Instead, much of the variation in species structure is driven by the loss/gain of species across the latitudinal gradient. Lock and Dam 19 does not appear to be a clear transition point in the river's fish community, although it may function as a meaningful barrier for particular species (e.g., invasive species) and warrant future attention from a management perspective.     

城市经济发展中公共图书馆服务社区读者的思考——以甘州区图书馆为例

随着张掖市甘州区社区经济的发展和建设水平的日益提高,社区文化在社区总体发展中显得日益重要。图书馆作为公益性服务单位和精神文明的窗口,担负着传递信息和公民终身教育的职责。在新形势下,要拓宽资源整合渠道,延伸服务领域,提高服务效能,努力增强公共图书馆的社会效益。分析了张掖市甘州区图书馆延伸社区服务的方式以及存在的问题,提出完善措施及今后努力的方向。     

基于MVC模式城市社区居民信息管理系统的设计

城市社区信息管理系统采用MVC模式进行设计,系统涉及到社区管理的各个方面,从居民基本信息管理,到外来人口统计等,都通过信息管理系统来得到高效的运行,系统实现了对整个居民信息管理的系统化、规范化和无纸化。整个系统使用户操作起来简便快捷,有效的减轻社区工作人员的劳动强度,提高工作效率与管理水平。     

基于边分类的SVM模型在社区发现中的研究

社区发现是复杂网络研究的重要内容,也是分析网络结构的重要途径。分析了社区发现研究中存在的问题,提出了一种基于边分类的SVM模型。通过边顶点相似度和边介数来表示边的特征,从而构造分类函数。利用LFR生成社区结构已知的人工网络,通过人工网络数据训练基于边分类的SVM模型,对分类函数的参数进行估计,利用训练模型对真实网络进行社区分类并通过标准化互信息(NMI)和整体准确度来评价分类效果。实验得到了较高的整体准确度和NMI值。实验表明基于边分类的SVM训练模型对真实网络数据的社区划分有较高的准确度,表明该方法是可行的。