基于多模态的音乐推荐系统

使用传统协同过滤的方式进行推荐往往会忽视音乐底层特征.通过将音乐的音频特征与歌词信息进行多模态融合,并将融合后的特征信息作为协同过滤推荐的补充,提出了一种基于多模态的音乐推荐系统.主要探讨了音频特征与歌词信息的提取,并在提取歌词信息时利用LDA主题模型进行特征降维.针对多模态融合问题,使用一种特征级联早融合法（EFFC）融合方式,并将多模态融合后的结果与单模态结果进行了比较.对于结果的推荐,以多模态特征信息为依据建立用户兴趣模型,并将该模型通过LSTM神经网络,以过滤与优化协同推荐的用户组.结果表明,基于多模态的音乐推荐系统将推荐结果的误差项平方和（SSE）由传统的2.009降至0.388 6,验证了该方法的有效性.     

Turning conflicts into cooperation? The role of adaptive learning and deliberation in managing natural resources conflicts in Nepal

ABSTRACT

Conflicts over natural resources are likely to escalate under changing socio-economic contexts and climate change. This paper tests the effectiveness of what we term Adaptive Learning and Deliberation (ALD) in understanding and addressing conflicts over the local management of forests and water, drawing on experimental work in Nepal. Based on a three-year action research project, the paper offers policy and practical insights on how complex and protracted conflicts can be addressed through the researcher-facilitated enquiry and deliberative processes that form the core of the ALD approach. The conflicts included in the study are a result of diverse environmental, political and economic factors. We analyze experimental practices in two sites, where our research team facilitated the ALD process, gathering evidence in relation to conflicting institutional issues, all of which was then fed into researcher-mediated and evidence-informed deliberations on conflict management. The analysis shows that the ALD process was helpful in rearranging local institutions to accommodate the interests of the conflicting groups and, to some extent, to challenge some of the underlying exclusionary provisions of forest and water institutions in Nepalese society. We also identify three key limitations of this approach – transaction costs, the need for strong research and facilitative capacity within the research team, and researchers’ engagement with the conflicting stakeholders.

Key policy insights

• Natural resource-based conflicts are intensifying in Nepal in recent years, due to heavy reliance of people on these resources for livelihoods, poor governance, and protection-oriented policies.

• Improved ways to facilitate cooperation among conflicting stakeholders are needed, as standard methods have often failed to address socio-environmental drivers of conflicts.

• The ALD approach can potentially help mitigate conflict and foster cooperation in natural resource management.

     

Graph-based synchronous collaborative mapping

Collaborative mapping is different from other collaborative editing systems because of spatial characterization and spatial dependence. Spatial operations on a mapping feature influence that feature and also the neighbouring features. Therefore, conflicts are likely to occur if common features or interrelated features are processed concurrently. For collaborative editing systems, the widely used multiversioning technique enables improved concurrency, which, however, is at the cost of synchronous data consistency. To maintain the synchronous consistency in collaborative mapping, a practical framework called the spatial operation dependence graph is proposed. The features are expressed as graph vertexes, while the dependences among the operations on features are expressed as graph edges. Through maintaining locking statuses of the vertexes, interdependent operations on features are arbitrated according to the temporal order, while independent operations can concurrently proceed. The utility of this framework is demonstrated through a prototype system, Co-Mapping.     

全球位置信息叠加协议与位置服务网技术研究进展与展望

随着信息技术的快速发展,当今世界步入信息爆发性增长的时代,互联网、物联网和传感网中的海量人类社会相关信息为解决城市治理和社会管理问题提供了新视角,如何将信息进行快速有效地整合和共享,是进一步提高城市智慧化管理能力面临的巨大挑战。近年来,在社会信息的位置泛在属性基础上提出的全球位置信息叠加协议与位置服务网技术,将位置作为信息横向关联的结点,为打通各领域系统间的壁垒,实现跨网、跨平台、跨系统、跨位置语言的海量异构信息自动汇集、融合计算与智能服务,提升社会化信息共享水平和效率,推动智慧地球和智能社会建设提供了支持。本文着重介绍了目前国内外关于全球位置信息叠加协议与位置服务网技术在立体剖分与位置编码、位置编码映射、地理实体编码、泛在位置信息叠加协议、高精度位置时空框架与位置感知的信息融合技术等各项关键技术的前沿进展。并进一步总结了全球位置服务网在行业中的示范应用,最后对全球位置服务网在技术和应用上的未来研究方向进行了讨论,以期望为未来全球位置服务网的发展提供参考。     

刺参（Apostichopus japonicus）幼参对高温和低盐协同胁迫的生理响应:生长和诱导型热休克蛋白基因表达

全球气候变化背景下的极端高温和强降雨频发天气导致的养殖池塘持续性的高温和低盐环境给刺参池塘养殖带来了严峻的挑战。为了研究刺参对高温和低盐环境的生理响应,本实验分析了高温和低盐协同胁迫对刺参幼参生长及诱导型hsp70和hsp90基因表达的影响。实验设置4个温度梯度（16℃、20℃、24℃和28℃）和3个盐度梯度（22、27和32）,共12个不同的胁迫组。经过40 d的长期胁迫,研究发现,随着温度的升高和盐度的降低,幼参的体重增加量减少,并在高温、低盐组出现体重负增长。长期胁迫提高了刺参幼参的诱导型hsp70和hsp90基因表达量,从而在一定程度上增强了对极端天气的抵抗能力。同时,高温下盐度22处理组刺参幼参诱导型hsp70和hsp90基因的表达量较盐度27条件下低。双因子方差分析结果显示,高温和低盐对幼参特定生长率和诱导型hsp70和hsp90基因表达量的交互作用均不显著,并且盐度对诱导型hsp70和hsp90基因表达量没有显著的影响。因此,相较低盐,高温对刺参幼参的影响更大,可作为刺参育种的选择压力。在高温期应该采取有效措施降低海水温度,并防止极端天气造成的养殖水体温度分层。该研究丰富了刺参生理生态学理论,可为刺参良种选育工作提供思路,并为指导极端天气下刺参生产实践活动提供一定的理论依据。     

驱动山区农业产业遥感监测的GMG协同系统建设与应用

由于喀斯特山区地貌的支离破碎、地形的高低起伏及气候的多云与雾,难以通过常规遥感手段实现大范围农作物的快速监测,针对该情况,本文研发了内外业一体化的GMG协同系统。该系统利用GPS的精准定位、多媒体的丰富语义及GIS的数据处理优势,多向驱动贵州山区农业产业遥感监测。实践表明,该系统快速实现了省域监测中解译样本的多样性,有效弥补了局部区域内遥感监测的局限性,同时检验了解译数据的准确性。     

勘察设计协同过程中存在的基本问题及解决途径探析

通过对勘察设计领域内的协同设计过程的分析,对存在的数据接口问题、数据复用问题进行了阐述,同时,针对这些问题提出了技术建议和解决方法。     

Multidimensional measurement of poverty and its spatio-temporal dynamics in China from the perspective of development geography

Exploring the spatio-temporal dynamics of poverty is important for research on sustainable poverty reduction in China.Based on the perspective of development geography,this paper proposes a panel vector autoregressive(PVAR)model that combines the human development approach with the global indicator framework for Sustainable Development Goals(SDGs)to identify the poverty-causing and the poverty-reducing factors in China.The aim is to measure the multidimensional poverty index(MPI)of China's provinces from 2007 to 2017,and use the exploratory spatio-temporal data analysis(ESTDA)method to reveal the characteristics of the spatio-temporal dynamics of multidimensional poverty.The results show the following:(1)The poverty-causing factors in China include the high social gross de-pendency ratio and crop-to-disaster ratio,and the poverty-reducing factors include the high per capita GDP,per capita social security expenditure,per capita public health expenditure,number of hospitals per 10,000 people,rate of participation in the new rural cooperative medical scheme,vegetation coverage,per capita education expenditure,number of universi-ties,per capita research and development(R&D)expenditure,and funding per capita for cultural undertakings.(2)From 2007 to 2017,provincial income poverty(IP),health poverty(HP),cultural poverty(CP),and multidimensional poverty have been significantly reduced in China,and the overall national poverty has dropped by 5.67％annually.there is a differentia-tion in poverty along different dimensions in certain provinces.(3)During the study period,the local spatial pattern of multidimensional poverty between provinces showed strong spatial dynamics,and a trend of increase from the eastern to the central and western regions was noted.The MPI among provinces exhibited a strong spatial dependence over time to form a pattern of decrease from northwestern and northeastern China to the surrounding areas.(4)The spatio-temporal networks of multidimensional poverty in adjacent provinces were mainly negatively correlated,with only Shaanxi and Henan,Shaanxi and Ningxia,Qinghai and Gansu,Hubei and Anhui,Sichuan and Guizhou,and Hainan and Guangdong forming spa-tially strong cooperative poverty reduction relationships.These results have important refer-ence value for the implementation of China's poVerty alleviation strategy.     

“本地和非本地”的北京企业联系：时空格局与影响因素

区域的中心—腹地关系一直是区域与城市发展领域关注的经典内容。全球化和信息化的推进产生了“流空间”和“中心流理论”,为现代城市间关系的研究提供了理论支撑,企业联系流反映的城市间关系受到学者的广泛关注,但多数研究聚焦在“整体空间”的城市网络特征表述,少有从特定问题出发,在更广阔的“非本地”范围研究中心城市与腹地的关系。论文基于1990—2018年县域尺度的企业流数据,研究北京与河北(本地)和京津冀区域外(非本地)的企业联系的时空格局和影响因素,从企业联系视角为北京与河北的经济落差提供解释,主要结论如下： ① 1990—2018年北京的本地性企业联系较弱,企业联系较强的县市主要集中在非本地,尤其是长三角等地区,非本地联系是北京企业联系的主要形式。从投资金额来看,北京与本地联系最强的行业为制造业,与非本地(主要是长三角)联系最强的行业为租赁和商务服务业;从建立分支数来看,北京与本地和非本地联系最强的行业都为金融业。② 北京的本地企业联系主要受本地的经营环境(金融环境、政府干预、政策环境)和区位条件(地理邻近)的影响,非本地企业联系主要受关键资源(政治资源、科技投入)、经营环境(经济开放)和市场潜力(消费水平)的影响。河北综合承载力较弱、与首都产业链不协调和不匹配等问题给北京与之建立本地联系上带来了一定的阻力,使北京倾向于与具有更优越市场潜力和经营环境的城市建立跨区域的非本地联系。     

协同创新对企业创新差距的影响机制——以高技术产业为例

以高技术产业数据为例,深入研究协同创新对企业创新差距的影响机制,借助面板回归模型、面板门槛模型和面板向量自回归模型,分析协同创新与企业创新差距之间的互动关系。研究结果表明：协同创新对企业创新差距的作用总体并不显著;协同创新对企业创新差距的作用存在区域异质性;协同创新水平较高时有利于缩小企业创新差距;创新差距较大时,协同创新能够缩小企业创新差距;协同创新与企业创新差距存在互动关系。