Regional innovation strategies in the knowledge-based economy

This paper aims to examine major regional innovation strategies for regional development in the knowledge-based economy in Korea. Innovation policy and Regional Innovation Systems (RIS) are becoming important issues in Korea in both national and regional perspectives. These issues have been evolving since the mid 1990s, especially since the financial crisis of 1997 in Korea. Five major policy issues for promoting innovation and regional competitiveness are examined in this paper. The major policy issues are: promoting region-specific clustering; building habitats for innovation and entrepreneurship; collective learning processes and innovation networks; building a stock of social capital; and promoting local and global networks. Several strategies are suggested for each of these policy issues. These major policy issues and related strategies for regional innovation and competitiveness can be regarded as common issues required for the successful development of RIS. Taking these basic policy issues into account, each region can develop their own region-specific RIS, with appropriate consideration of distinctive regional characteristics such as the level of development, industrial specialization, local labor market, local business climate, local financial system, etc.     

Promoting Lifelong Professional Development in Geography Education: International Perspectives on Developing the Scholarship of Teaching in Higher Education in the Twenty‐First Century*

Whereas the global march towards the professional development of geography teachers in schools was, perhaps, the major achievement in geography education of the twentieth century, promoting the professional development of faculty teaching geography in higher education is one of the major challenges that faces us in geography education for the twenty‐first century. What little educational development is currently provided for faculty teaching in universities is largely generic. There is a need for a disciplinary specific component in educational development, both for the initial training of faculty and for their continuing professional development. This article explores how the professional development of faculty is related to the international debate about developing the scholarship of teaching and learning in higher education. It reviews some recent initiatives, particularly from the United States and the United Kingdom, and examines the role of national associations and international networks in supporting faculty teaching geography in higher education. Some of the challenges involved in professionalizing teaching geography in higher education are discussed. The article ends with some suggestions for promoting lifelong professional development in geography in higher education nationally and internationally.     

Using an Artificial Intelligence Method to Recognize Seismic Precursory Anomaly (SPA)

In this work,we seek for a new way to recognize and analyze the Seismic Precursory Anomaly(SPA)quantitatively.For further study of the mechanism of SPA,an artificial intelligence method,the Machine Discovery(MD)method,is proposed to process the Seismic Precursor Observation(SPO)data and to fit an analytic function to them.The analytic function is used as a criterion to identify the SPA.The algorithm of MD is worked out based on the theory of Recursive Function(RF).The data of Amplitude Factor of Solid Tide(7 value)are processed by the proposed method.We find that the subsequent data fit to the extrapolation of the discovered function during the quiet period,whereas they deviate from it before a strong earthquake.Finally,the feasibility of quantitative analysis of the SPA is testified.     

Restructuring of Australian higher education: information technology in geography teaching and learning

Australian higher education has experienced substantial change since the early 1980s, with a transition to mass higher education, closer alignment to national political objectives, increasing Federal Government control, a growing emphasis on ‘quality’ and ‘value for money’, and organisational restructuring with manifestations ranging from the creation of the Unified National System to the merger or closure of individual departments. At the same time, evolving conceptions of learning and new patterns of demand, with growing emphasis on lifelong learning and flexible access to education, pose new challenges for educators. Now, the policies of the new Federal Coalition Government are likely to engender further change, with growing differentiation of universities a probable outcome. Information technology has so far played a relatively peripheral role in teaching and learning in higher education, but important changes there, including rapid shifts in the relationship between cost and computing power, and the explosive emergence of the World Wide Web, are now providing the basis on which information technology can become increasingly central to teaching and learning. Higher education is close to the limits of incremental adjustment to pressures for change and it seems likely that it will experience major restructuring, with its eventual reconstruction around the capabilities of information technology. Geography has already been much affected by restructuring pressures, including above‐average growth in teaching loads and threats to its disciplinary identity because of internal reorganisation of many universities. Despite a tradition of innovative, high‐quality teaching, geography has not generally been a leader in the development and application of information technology. Nevertheless, a case study of an experimental use of the World Wide Web provides an example of directions that teaching and learning in the discipline might take.     

基于多源数据和集成学习的城市住宅地价分布模拟——以武汉市为例

精准刻画城市住宅地价分布特征,对于科学引导城市空间布局规划、有效实现城市精明增长等具有重要意义。而城市住宅地价与其潜在影响因素之间的复杂非线性关系,给地价分布精细模拟带来了挑战。论文旨在探索基于地理大数据和集成学习的城市住宅地价分布模拟方法体系,以满足快速、精准监测地价动态变化的需要。选取武汉市为典型区,以住宅用地交易样点、兴趣点(points of interest, POI)和夜间灯光影像为数据源,以500 m分辨率网格为估价单元,提取POI核密度和夜间灯光强度作为住宅地价预测变量,采用机器学习算法和bagging、stacking集成方法构建住宅地价预测模型,并对比分析其精度。研究发现：① 单个机器学习算法中,支持向量回归(support vector regression, SVR)预测精度最高,接下来依次是k最近邻算法(k-nearest neighbor algorithm, k-NN)、高斯过程回归(Gaussian process regression, GPR)和BP神经网络(back propagation neural networks, BP-NN);② 在提升单个算法预测精度方面,stacking方法的性能优于bagging方法,使用stacking集成SVR和k-NN的地价预测模型精度最高,其平均绝对百分误差仅为8.29%,拟合优度R²达0.814;③ 基于论文所构建模型生成的城市住宅地价分布图能有效表征价格圈层分布特征和局部奇异性。研究结果可为城市住宅地价评估提供新的思路和方法借鉴。     

基于RNN的空气污染时空预报模型研究

针对空气污染物时间序列中包含缺失值以及现有时间序列预报模型缺乏对时序特征状态建模的问题,该文构建了基于缺失值处理算法和RNN(循环神经网络)的时空预报框架。对空气污染物时序数据设计了3种缺失值处理算法(前向递补、均值替代和权重衰减),用缺失标签和缺失时长对缺失值建模,并在此基础上搭建含有全连接层与LSTM层的深度循环神经网络(DRNN)用于时空预报。使用深度全连接神经网络(DFNN)作为DRNN的对照,用京津冀区域的空气质量和气象数据训练模型,并比较不同模型的预测精度。通过实验,比较了3种缺失值处理方法的效果,结果表明,LSTM在空气污染时空序列预测上的表现优于传统的全连接神经网络层,证实了提出的基于深度学习的时空预报框架的有效性。     

一种基于半监督学习的地理加权回归方法

地理加权回归方法在小样本数据下回归分析精度往往不高。半监督学习是一种利用未标记样本参与训练的机器学习方法,可以有效地提升少量有标记样本的学习性能。基于此本文提出了一种基于半监督学习的地理加权回归方法,其核心思想是利用有标记样本建立回归模型来训练未标记样本,再选择置信度高的结果扩充有标记样本,不断训练,以提高回归性能。本文采用模拟数据和真实数据进行试验,以均方误差提升百分比作为性能评价指标,将SSLGWR与GWR、COREG对比分析。模拟数据试验中,SSLGWR在3种不同配置下性能分别提升了39.66%、11.92%和0.94%。真实数据试验中,SSLGWR在3种不同配置下性能分别提升了8.94%、3.36%和5.87%。SSLGWR结果均显著优于GWR和COGWR。试验证明,半监督学习方法能利用未标记数据提升地理加权回归模型的性能,特别是在有标记样本数量较少时作用显著。     

基于稀疏表示和字典学习的全色与多光谱影像融合

Pansharpening方法通过融合多光谱影像的光谱信息和全色影像的空间细节信息来得到高分辨多光谱影像。然而传统的Pansharpening方法易导致产生光谱扭曲和空间信息丢失现象。受到影像稀疏表示超分重建理论启发,本文提出了一种新的基于稀疏表示和字典学习的Pansharpening方法。该方法以影像的高频特征作为训练样本,通过字典学习的方法来获取高低分辨率影像字典,使用正交匹配追踪算法求解出影像的稀疏表示系数,最终通过高分辨影像字典与稀疏系数相乘得到融合影像。实验结果表明:本文提出的方法能很好地保持遥感影像的光谱信息和空间细节信息。     

深度学习的半监督遥感图像检索

遥感图像数据的海量性、多样性和复杂性等特点对遥感图像检索的速度和精度提出了更高的要求,其中特征提取是影响遥感图像检索效果的关键。本文方法首先对遥感图像进行预处理,然后基于稀疏自动编码的方法在大量未标注的遥感图像上进行特征学习得到特征字典,基于卷积神经网络的思想,使用训练出来的特征字典对遥感图像进行卷积和池化得到每幅图像的特征图;接下来使用特征图训练Softmax分类器;最后对待检索图像分类,在同一类别中计算特征间的距离,进而实现遥感图像的检索。实验结果表明,该方法能够有效提高遥感图像检索的速度和准确度。