党参多糖对小鼠学习记忆作用研究

研究党参多糖对东莨菪碱所致记忆获得障碍模型小鼠(跳台法和Y水迷宫法)、亚硝酸钠所致记忆巩固障碍模型小鼠(跳台法)和40%乙醇所致记忆再现障碍模型小鼠(暗箱法)学习记忆的影响。结果表明:党参多糖给药组均能不同程度缩短东莨菪碱所致记忆获得障碍模型小鼠的触电潜伏期(P0.01或P0.05),提高空间记忆能力,缩短寻觅平台的潜伏期(P0.01);能延长亚硝酸钠所致记忆巩固障碍模型和40%乙醇所致记忆再现障碍模型小鼠的记忆潜伏期(P0.01或P0.05),减少错误次数(P0.01)。说明党参多糖具有改善小鼠学习记忆能力的作用。     

基于小波变换的LMSV模型对人民币汇率波动的研究

将小波引入到LMSV模型波动长记忆性的估计与检验中,提出了基于小波变换的LMSV模型波动长记忆性的伪极大似然估计法和波动长记忆性的检验方法,并对各汇率波动序列长记忆效应的大小程度进行了验证.结果表明各汇率波动序列存在长记忆效应,人民币对美元的汇率波动序列受历史信息的影响程度最高.     

The mycological properties of medieval culture layers as a form of soil ‘biological memory’ about urbanization

Background, aim, and scope  The mycological properties of soil can offer information about ancient human–landscape interaction, including urbanization. This preposition has been confirmed in our study of the habitation deposits in the medieval Russian settlements (eighth to fourteenth centuries A.D.). Materials and methods  The mycobiota of profiles of anthropogenically transformed soil of excavated medieval settlements were examined in different climate conditions in the European part of Russia. The fungal biomass and biomass structure were evaluated using luminescent microscopy. The isolation and enumeration of microfungi were performed using the method of serial dilutions of soil samples and plating them out on the number of solid media. The isolation of keratinolytic microfungi was performed by hear-bite technique. Results  It was established that in ancient urban soils, the mycobiota may have properties different from those of zone fungal communities. The examined cultural layers of ancient settlements differ from the horizons of the surrounding natural soils because of a bigger rate of fungal spores in fungal biomass, more mosaic distribution of microfungal communities, different species composition and dominant microfungal species, increased incidence of some ecological, and trophic fungal groups (for example, keratinophilic, potentially pathogenic microfungi). Discussion  The mycological characteristics of anthropogenic deposits in the excavated medieval settlements were found to be mainly similar to mycological properties of modern urban soils. Conclusions  These properties of habitation deposits can be interpreted as a kind of “soil mycological memory” of the ancient urban impact. Recommendations and perspectives  Soil mycological characteristics could be used in paleoecological reconstructions and biomonitoring of urban impacts     

基于记忆梯度混合遗传算法的灌区水资源优化配置

针对灌区水资源优化配置模型中目标函数高度非线性的特点,该文提出基于记忆梯度混合遗传算法用于灌区水资源优化配置,并提出了有条件的随机生成的初始种群生成方式来处理线性约束条件的策略.实例研究表明算法用于灌区水资源优化配置比基本遗传算法计算精度有了明显提高.     

限制性内切酶在PCR缓冲液中酶切活性的研究(英文)

选取6 种常用的限制性内切酶(Acc I、Ban II、EcoR I、Hind III、Sac I、Sca I)用于检测其直接在PCR 缓冲液中完全酶切DNA 的活性。结果表明:在PCR 缓冲液中,需要加入MgCl2作为激活因子的前提条件下,除了EcoR I 表现出星状活性外,其他酶均能完全酶切无甲基化λDNA,其结果与在酶生产厂商推荐的反应缓冲液中进行的酶切反应结果一致;在完成PCR 扩增的PCR 缓冲液中,所有酶(包括RcoR I)均能功能正常地对PCR 产物完全酶切,表明可在PCR 扩增完毕后的反应体系中直接加入限制性内切酶进行酶切反应,无需对PCR 产物进行预先纯化;镁离子浓度在2.5 mmol·L-1至10 mmol·L-1范围内对PCR 缓冲液中的限制性内切酶的活性无显著影响。上述方法可用于限制性片段长度多态(RFLP)以及单链构像多态(SSCP)研究中,但能否用于分子克隆有待进一步实验。图3 表1 参7。     

递归神经网络在降雨量预测中的应用研究

递归神经网络(RNN)模型近年来在许多任务上表现出了优良的性能。运用具有长短期记忆(LSTM)单元的递归神经网络构建模型和通过时间反向传播(BPTT)算法更新网络权重解决长期降雨量的预测问题,较好地解决了高维数、非线性和局部极小问题。选取了前馈神经网络模型(FNN)、小波神经网络(WNN)模型和整合移动平均自回归(ARIMA)模型3种模型进行验证比较。仿真结果表明,递归神经网络模型优于其他模型,训练结果与实际值接近,预测精度较高。预测结果为农业用水管理、合理制定灌溉制度提供了重要的科学依据。     

The buffer effect and the selection of protected areas for waterbirds in Britain

The buffer effect predicts that where the reproductive success and survivorship of a species vary between potential habitats, sites will be sequentially filled according to a preference hierarchy. Once favoured sites reach saturation, numbers on the less-suitable/poorer quality sites will show a greater rate of increase compared with those on favoured sites. Supporting evidence for a buffer effect is readily available in the literature for many species, although this is generally restricted to small-scale analyses. In this paper we test for a buffer effect on a national scale for 19 species of waterbirds regularly wintering in the UK for which populations have increased nationally. The results provide little support for the effect, with only four species showing significant negative correlations. Nonetheless, a number of factors are likely to confound the occurrence/identification of a buffer effect for these species, including site area and data limitations. By contrast, for the majority of these 19 species, those sites where initial population totals are largest are also those with the fastest rates of population increase. Encouragingly, these sites are, therefore, more likely to be classified and managed as Special Protection Areas (SPAs) or Ramsar sites using the current numerical criteria for identifying such conservation areas.     

采用改进的SqueezeNet模型识别多类叶片病害

为实现作物叶片病害的准确识别,该研究以PlantVillage工程开源数据库中14种作物38类叶片为研究对象,从网络规模小型化和计算过程轻量化需求的角度出发,对经典轻量级卷积神经网络SqueezeNet提出改进措施,包括修改最后一层卷积层的输出、删除经典模型中的后3个fire模块并修改fire模块5的参数、调节fire模块中expand层中1×1和3×3的卷积核数目的比例、移动部分fire模块在模型中的位置等措施,共获取5种改进的病害叶片检测模型,并运用迁移学习和随机梯度下降算法进行训练。试验结果表明,在不过多损失网络性能的前提下,改进后5种模型的参数内存需求及模型计算量均呈现大幅减小,模型收敛迅速,其中最优模型参数内存需求仅为0.62MB,模型运算量仅为111MFLOPs,其平均准确率达到98.13%,平均查全率达到98.09%,平均查准率达到97.62%,在与已有相关研究的对比中表现出较高的性价比。该研究提出的改进模型在大幅减少参数内存要求和计算量的同时使模型性能保持在一个较高的水平,较好地平衡了这3项指标,适合将模型部署在移动终端等嵌入式资源受限设备上,有助于实现对作物病害的实时准确识别。     

采用改进长短时记忆神经网络的水产养殖溶解氧预测模型

为了精确预测水产养殖溶解氧变化趋势,该研究提出了基于K-means聚类和改进粒子群优化(Improved Particle Swarm Optimization,IPSO)的长短时记忆(Long Short-term Memory,LSTM)神经网络预测模型。根据环境因子间的相似度,应用改进的K-means聚类算法将环境数据划分为若干类。在此基础上,基于LSTM神经网络算法构建改进的水产养殖溶解氧预测模型,并引入改进粒子群优化算法对模型参数进行优化,以减少经验选取参数的盲目性。在不同天气状况下利用该模型对溶解氧进行预测。试验结果表明,在良好天气情况下,该模型预测误差曲线波动较小,预测精度更高。当天气发生突变时,溶解氧预测模型评价指标平均绝对百分误差、均方根误差、平均绝对误差和纳什系数分别为0.129 5、0.645 3、0.461 3和0.902 2。该模型一定程度改善了天气突变状况下的数据缺失、鲁棒性差等问题。     

景观改造中对历史遗址的保护研究——以河南郑州市紫荆山人民广场改造为例

河南省郑州市紫荆山人民广场为一占地约2.2 hm2的街旁绿地,郑州市政府2011年将其规划为人民广场。从遗址保护和文化传承的角度出发,通过对现状的分析和现存问题的讨论并重新对该广场进行设计改造,旨在传承发扬历史文化,为大众提供一个功能性与纪念性于一身的开放型城市街旁绿地。