GA-BP网络模型在GPS似大地水准面精化中的应用

针对误差反向传播（BP）算法训练速度慢和易于陷入局部最小值的缺点,提出了利用遗传算法（GA）的全局寻优性,结合GA和BP的各自优点,分析和建立了进化神经网络（GA-BP）模型,并将该模型应用于似大地水准面模型精化。最后以南方某市E级GPS控制网高程数据为例,进行BP和GA-BP模型的对比实验,通过对内、外符合精度及MAPE（平均绝对误差百分比）指标分析,验证了该方法的可行性。     

常德1631年大震事件的多场点群发式构造模型

根据对常德1631年地震事件的史料鉴别,尤其是江陵为典型七度震害这一事实,运用震声传播方向和地光物理场以及地震烈度衰减关系,拟合推断此地震事件的序列主要由安乡北6 3/4级地震、常德北太阳山7级地震、澧县6级地震、大庸6级地震和澧县—安乡6 1/2级地震组成,相应的地震地质背景是江汉—洞庭地区晚更新世以来的剩余形变呈四象限分布。若考虑地震-构造相关分布为“Z”字型,则该地震事件的主控构造——太阳山断裂带东断裂具左旋走滑的力学机制。根据区域地震—构造相关发展过程,常德大震事件可视为1668年郯城8 1/2级地震的前震序列组成部分。     

1963、1976年两次不同ELNino事件的差异以及我国旱涝和台风的对比分析

本文通过对1963、1976年两次ELNino事件的分析指出:两次ELNino事件在很多方面较为一致,其主要差别是ELNino开始的地理位置(主要是经度位置)和海温分布不同。同时,又分析了两类不同型式的ELNino年对我国干旱、雨涝和登陆台风的影响。     

卫星气候数据集的应用研究与发展分析

卫星气候数据集是卫星气候研究的基础。在规范卫星气候数据集基本概念的基础上,针对现有基本气候数据集（FCDR）和专题气候数据集（TCDR）的分类方式,无法反映卫星气候数据特点的问题,认为应将专题气候数据集进一步划分为单一遥感仪器专题气候数据集、多种遥感仪器融合专题气候数据集及卫星与多源资料融合专题气候数据集等几类。这种分类方法便于用户更好地了解和使用卫星气候数据。然后,重点围绕基本气候变量和基本卫星气候变量含义、卫星气候数据集生产规范、国内外主要卫星气候数据生产计划等方面,综述了卫星气候数据集建设及规范化生产已取得的最新研究进展。在此基础上,分析了卫星气候数据集建设和应用中存在的主要问题,展望了卫星气候数据集发展,同时对我国卫星气候数据集建设提出具体建议。     

Funding low-carbon investments in the absence of a carbon tax

Introducing a carbon tax is difficult, partly because it suggests that current generations have to make sacrifices for the benefit of future generations. However, the climate change externality could be corrected without such a sacrifice. It is possible to set a carbon value, and use it to create ‘carbon certificates’ that can be accepted as part of commercial banks’ legal reserves. These certificates can be distributed to low-carbon projects, and be exchanged by investors against concessional loans, reducing capital costs for low-carbon projects. As the issuance of carbon certificates would increase the quantity of money, it will either lead to accelerated inflation or induce the Central Bank to raise interest rates. Low-carbon projects will thus have access to cheaper loans at the expense of either ‘regular’ investors (in case of higher interest rates) or of lenders and depositors (in case of accelerated inflation). Within this scheme, mitigation expenditures are compensated by a reduction in regular investments, so that immediate consumption is maintained. It uses future generation wealth to pay for a hedge against climate change. This framework is not as efficient as a carbon tax but is politically easier to implement and represents an interesting step in the trajectory towards a low-carbon economy.     

基于粗糙集与人工神经网络的变压器故障诊断

根据电力变压器故障诊断问题,提出了基于粗糙集与人工神经网络的变压器故障诊断模型,分析了该模型的实现步骤.采用Kohonen网络对连续属性值进行离散化,应用粗糙集理论对特征参数进行属性约简,并把约简结果生成规则作为BP网络的输入.仿真结果表明,把经过粗糙集理论预处理过的数据送入BP网络训练,提高了学习速度和故障诊断正确率,减少了训练时间.     

Climate Change of 4°C Global Warming above Pre-industrial Levels

Using a set of numerical experiments from 39 CMIP5 climate models, we project the emergence time for 4°C global warming with respect to pre-industrial levels and associated climate changes under the RCP8.5 greenhouse gas concentration scenario. Results show that, according to the 39 models, the median year in which 4°C global warming will occur is 2084. Based on the median results of models that project a 4°C global warming by 2100, land areas will generally exhibit stronger warming than the oceans annually and seasonally, and the strongest enhancement occurs in the Arctic, with the exception of the summer season. Change signals for temperature go outside its natural internal variabilities globally, and the signal-to-noise ratio averages 9.6 for the annual mean and ranges from 6.3 to 7.2 for the seasonal mean over the globe, with the greatest values appearing at low latitudes because of low noise. Decreased precipitation generally occurs in the subtropics, whilst increased precipitation mainly appears at high latitudes. The precipitation changes in most of the high latitudes are greater than the background variability, and the global mean signal-to-noise ratio is 0.5 and ranges from 0.2 to 0.4 for the annual and seasonal means, respectively. Attention should be paid to limiting global warming to 1.5°C, in which case temperature and precipitation will experience a far more moderate change than the natural internal variability. Large inter-model disagreement appears at high latitudes for temperature changes and at mid and low latitudes for precipitation changes. Overall, the inter-model consistency is better for temperature than for precipitation.     

基于Item2Vec负采样优化的专题地图产品个性化推荐方法研究

建立适用于专题地图产品检索的用户偏好推荐模型是提高专题地图质量的有效方式之一,在专题地图产品推荐场景中,存在严重的内容冷启动和评论数据稀疏问题,现有的推荐算法无法为特定类用户推荐不同特征的专题地图产品,导致用户从专题地图中获取偏好信息受到限制。因此,本文构建基于负采样的连续词袋模型和基于Word2Vec的Item2Vec相结合的用户偏好推荐方法,用于专题地图产品推荐。① 计算用户行为日志文件中交互行为数据的隐性评分,以代替专题地图推荐信息流场景中稀疏的用户评论数据;② 基于负采样的连续词袋模型提取目标专题地图的前后地图序列感知特征信息,通过控制正负样本比例为1:2,提升目标专题地图潜在评分的预测精度;③ 通过Item2Vec将带有用户行为特征信息的专题地图映射到向量空间,计算用户对专题地图的相似度矩阵,根据用户偏好程度完成推荐。在构建的专题地图评分实验数据集Thematic CMaps和4个公开验证数据集MovieLens上的测试结果表明:与LFM、Personal Rank、Content Based和SVD 4种传统推荐算法相比,本文所提方法可有效提高潜在评分的预测精度,推荐性能最高达到27.85%;与以霍夫曼采样方式的Item2Vec基础方法和YouTubeNet 2种神经网络推荐算法相比,评分预测精度有一定提高,且推荐性能不断提升,最高达到2.97%和5.78%。以经典算法奇异值分解（SVD）为例,将MovieLens-20M数据集切分后,在数据量不断增大的数据子集中,本文所用方法的评分预测精度和性能均优于SVD方法。     

Source Evolution and Its Relationship to Climate Change Since the Middle–Late Pleistocene in Laizhou Bay,China

Through the geochemical analysis of two hundred-meters cores KD4 and ZK3 from Laizhou Bay,in this study,we determined the distribution law and controlling factors of the geochemical elements.We analyzed 24 elements with respect to their R factors and major principal components,which were combined with the source discrimination functions DFCr/Th and DFCa/Al to obtain the sediment source index and its variation with depth for this area.A comparison of the changes in climate indicators suggests a clear correlation between the source and climate changes.The results show that the Yellow River and surrounding short-term rivers are the main sediment sources in this area.The PC3 of the KD4 core and PC2 of the ZK3 core(e.g.,CaO,MnO,SiO2)exhibit significant variations and reflect the relative contributions of Yellow River sources.The deposition process can be divided into six stages:in Phase I(MIS 5c–MIS 5a),the Yellow River formed,and the composition of the Yellow River had a greater influence on the sedimentary composition of the study area.In Phase II(MIS 5a–MIS 3),the sediment sources of the Yellow River and the short-term streams in this area were wavering,with the sediments derived from short-term rivers playing a more important role.In Phase III(MIS 3),with a sharp drop in temperature,the study area was in the process of retreat,and the sediment source changed from the Yellow River to short-term rivers,after which the Yellow River source material remained the main sediment source for the region.A similar process occurred three more times in Phase IV(MIS 3–MIS 2),Phase V(MIS 2–MIS 1),and Phase VI(MIS 1).With changes in climate,especially during alternating sea-land phases,the sediment source varied in marine-terrestrial-marine phases,and the changes are observed as Yellow River source-surrounding provenance-Yellow River source.However,this process of change is not synchronized with the sea-land strata alternation.     

HISTORICAL DESERTIFICATION PROCESS IN HEXI CORRIDOR, CHINA

1IN TR O D U CTIO NH istoricdaelsertificatiisonnotonlythekeyissueon theinteractiobnetweennatureandhuman proces,sbutalsothemost challenginigssueinthecomplex reltaionshipbetweenhuman andland.Previousl,ymostChinesesci-entistisnthisfieldcarrieoduttheiresearch…