PALEOCLIMATE: Glacial Climate Instability

Throughout the last glacial period, rapid climatic changes called Dansgaard-Oeschger (D-O) events occurred in the Northern Hemisphere. As Labeyrie discusses in his Perspective, these events are ideal targets for testing our understanding of climate change and developing climatic change models. Important steps toward understanding D-O events, particularly regarding the role of the low latitudes, are now reported by Hughen et al. and Peterson et al.     

Milankovitch radiation variations: a quantitative evaluation

A quantitative determination of changes in the surface temperature caused by variations in insolation calculated by Milankovitch has been made through the use of the thermodynamic model of Adem. Under extreme conditions, mean coolings of 3.1 degrees and 2.7 degrees C, respectively, at latitudes 25 degrees and 65 degrees N are obtained for Milankovitch radiation cycles. At the sensitive latitude 65 degrees N, a mean cooling below the present temperature for each of the times of radiation minimum is only 1.4 degrees C. This result indicates that the Milankovitch effect is rather small to have triggered glacial climates.     

PALEOCLIMATE: The Amazon Reveals Its Secrets--Partly

The role of the tropics in global climate change during glacial cycles is hotly debated in paleoclimate cycles today. Records from South America have not provided a clear picture of tropical climate change. In his Perspective, Betancourt highlights the study by Maslin and Burns, who have deduced the outflow of the Amazon over the past 14,000 years. This may serve as a proxy that integrates hydrology over the entire South American tropics, although the record must be interpreted cautiously because factors other than rainfall may contribute to the variability in outflow.     

PALEOCLIMATE: Taking the Temperature of Past Ocean Surfaces

Accurate reconstructions of past climatic conditions on Earth are important for modeling future climate change. Researchers use climate "proxies" from deep-sea sediments, corals, or ice cores to reconstruct these past climatic conditions, but as Nürnberg explains in this Perspective, different techniques often give different answers. A relatively new technique, Mg/Ca paleothermometry, is a particularly promising tool, especially in conjunction with oxygen isotope measurements on the same planktonic foraminiferal specimens. The power of this method is exemplified by the study by Lea et al.     

PALEOCLIMATE: An Appealing Snowball Earth That's Still Hard to Swallow

The startling claim that Earth has frozen over from pole to pole for millions of years at a time, a hypothesis widely known as the "snowball Earth" scenario, has intrigued many earth scientists but as yet convinced few. But the theory, which would explain some previously inexplicable observations from the Neoproterozoic era 600 million to 700 million years ago, is gaining adherents. Last year, for example, most researchers agreed that one part of the sweeping hypothesis--the claim that glaciers once flowed into ice-covered tropical seas--is correct, even though this idea had been rebuffed for more than 30 years.     

MSP问题的一个求解算法

利用改进的数据包络分析模型测度2000年以来30个省（市）跨期可比的生态效率,考察其时空演变特征, 基于空间杜宾模型分析各类因素影响生态效率的直接效应和空间溢出效应。研究发现：区域生态效率呈空间集聚态势,具有路径依赖特征;专利授权总数和资本劳动比以及外商直接投资均具有积极的直接效应,且对邻近区域具有正向空间溢出效应;城镇化的各类效应均不显著,以原燃料和动力价格指数代表的价格机制、企业研发人力投入占比反而对本地生态效率有显著负向影响。建议：以生态效率为核心建立区域考核约束机制,在提升城镇化质量、改革资源价格形成机制、优化创新投入结构等的同时,还需重视各类机制和结构的协同创新,进行系统化设计和差异化改革。     

解约束优化问题的一种信赖域方法

提出了一种解约束光滑优化问题的依赖域方法，并给出了算法的全局收敛性。     

不良贷款率对银行业影响的统计关系检验

在5%的显著性水平下,不良贷款率对资本利润率、净利差、资本充足率有单向格兰杰因果关系;而存贷比、流动性比率对不良贷款率有单向格兰杰因果关系;拨备覆盖率与不良贷款率在不同滞后期格兰杰因果关系不同。当不良贷款率上升时,商业银行不应通过承担更高的风险选择高贷款利率的贷款项目以实现其经营绩效;在系统性风险可控的前提下,监管部门和中央银行实施逆周期调整,可适当有选择性地提高容忍度,允许部分商业银行或业务条线调整对资本的计提标准,降低其顺周期性;商业银行应注重盈利性、流动性、安全性之间的平衡,理性对待不良贷款率上升。     

求解无约束极大极小问题的光滑化不精确牛顿算法

提出了求解无约束极大极小问题的光滑化不精确牛顿算法.该算法利用光滑凝聚函数近似不可微的极大值函数,从而得到目标函数的光滑近似,进而再利用不精确牛顿法求解光滑化后的可微的无约束优化问题.在一定的假设条件下,算法具有全局收敛性,初步的数值实验表明,算法是有效的.     

一种解决BP网络过拟合的学习算法

针对BP网络的过拟合问题,提出了一种基于验证误差的学习算法,该算法主要从构建最佳网络结构及获得最适训练次数两个角度进行了考虑.在学习过程中,通过不断检验验证误差及误差变化量,自动调整隐层节点数目,并适时停止网络训练.针对给定问题,该算法不仅能自动构建最佳网络结构和相应权系数,而且同时对网络的泛化能力进行了检验,从而有效避免了网络的过拟合.最后通过实验证明了算法的有效性.     

各向异性杆扭转问题的一种数值方法

将R-函数理论用于分析具有任意形状截面的各向异性杆弹性扭转问题.首先引入坐标变换将扭转问题的应力函数方程化成拉普拉斯算子方程,再利用拉普拉斯算子的基本解、边界方程及R-函数构造一个准Green函数,其满足了齐次边界条件.后再应用Green公式将各向异性杆弹性扭转问题的拉普拉斯算子方程转化为积分方程,应用R-函数理论选择适当的边界方程来消除积分方程的奇异性.数值算例表明,此方法是一种有效的数值方法.     

一种解决BP网络过拟合的学习算法

针对BP网络的过拟合问题,提出了一种基于验证误差的学习算法,该算法主要从构建最佳网络结构及获得最适训练次数两个角度进行了考虑.在学习过程中,通过不断检验验证误差及误差变化量,自动调整隐层节点数目,并适时停止网络训练.针对给定问题,该算法不仅能自动构建最佳网络结构和相应权系数,而且同时对网络的泛化能力进行了检验,从而有效避免了网络的过拟合.最后通过实验证明了算法的有效性.     

基于Memetic算法的多模式约束项目调度问题研究

为解决多模式资源约束项目调度问题(MRCPSP),在建立数学模型基础上,提出一种改进的Memetic算法对模型进行求解.该算法利用遗传算法进行全局搜索,采用针对活动执行模式的整数编码方式,基因的值表示活动的优先权和执行模式,每条染色体对应一个满足资源约束的项目调度方案,种群在交叉和变异后采用模拟退火算法进行局部优化.实例仿真结果表明了该算法的有效性和高效性.     

线性凸规划问题的一个势降不可行内点算法

基于 Tanabe-Todd-Ye 势函数提出了求解线性凸规划问题的一个势降不可行内点算法,分析了算法的复杂度,并证明了在每一次迭代之后算法使用的势函数至少有一个正常数的下降量,从而说明了算法的多项式复杂性。     

求解约束全局最优化问题的一个新的填充函数法

填充函数法是求解全局优化问题的一类有效方法,其关键之一在于构造一个叫作填充函数的辅助函数,针对一般约束优化问题,提出了一个新的填充函数.在适当的假设条件下,讨论了其良好的填充性质,并按其理论性质设计了一个合理的算法,而且理论上的证明表明该函数较其他多参数填充函数具有优越性.     

一种新的求解无约束优化问题的谱共轭梯度法

在修正的Wei-Yao-Liu共轭梯度法基础上,给出一种新的求解无约束优化问题的谱共轭梯度算法,该算法在强wolfe型线搜索下具有充分下降性和全局收敛性,数值实验结果表明该算法是有效的.